diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_cdist_backward_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_cdist_backward_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..a574de57255e84db15fb6b13e1e32d84bf58babc
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_cdist_backward_native.h
@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor & _cdist_backward_out(const at::Tensor & grad, const at::Tensor & x1, const at::Tensor & x2, double p, const at::Tensor & cdist, at::Tensor & out);
+TORCH_API at::Tensor _cdist_backward(const at::Tensor & grad, const at::Tensor & x1, const at::Tensor & x2, double p, const at::Tensor & cdist);
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_ctc_loss_backward_compositeexplicitautograd_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_ctc_loss_backward_compositeexplicitautograd_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..89308c624a535ec2aa0ad367a3c657bcc0a80143
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_ctc_loss_backward_compositeexplicitautograd_dispatch.h
@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API at::Tensor & _ctc_loss_backward_out(at::Tensor & out, const at::Tensor & grad, const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, const at::Tensor & neg_log_likelihood, const at::Tensor & log_alpha, int64_t blank, bool zero_infinity=false);
+TORCH_API at::Tensor & _ctc_loss_backward_outf(const at::Tensor & grad, const at::Tensor & log_probs, const at::Tensor & targets, at::IntArrayRef input_lengths, at::IntArrayRef target_lengths, const at::Tensor & neg_log_likelihood, const at::Tensor & log_alpha, int64_t blank, bool zero_infinity, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_cudnn_rnn_flatten_weight_ops.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_cudnn_rnn_flatten_weight_ops.h
new file mode 100644
index 0000000000000000000000000000000000000000..f6f797c36c43db92d391618949b2e26eed5c5fff
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_cudnn_rnn_flatten_weight_ops.h
@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _cudnn_rnn_flatten_weight {
+  using schema = at::Tensor (at::TensorList, int64_t, c10::SymInt, int64_t, c10::SymInt, c10::SymInt, int64_t, bool, bool);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_cudnn_rnn_flatten_weight")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_cudnn_rnn_flatten_weight(Tensor[] weight_arr, int weight_stride0, SymInt input_size, int mode, SymInt hidden_size, SymInt proj_size, int num_layers, bool batch_first, bool bidirectional) -> Tensor")
+  static at::Tensor call(at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional);
+};
+
+struct TORCH_API _cudnn_rnn_flatten_weight_out {
+  using schema = at::Tensor & (at::TensorList, int64_t, c10::SymInt, int64_t, c10::SymInt, c10::SymInt, int64_t, bool, bool, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_cudnn_rnn_flatten_weight")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_cudnn_rnn_flatten_weight.out(Tensor[] weight_arr, int weight_stride0, SymInt input_size, int mode, SymInt hidden_size, SymInt proj_size, int num_layers, bool batch_first, bool bidirectional, *, Tensor(a!) out) -> Tensor(a!)")
+  static at::Tensor & call(at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList weight_arr, int64_t weight_stride0, c10::SymInt input_size, int64_t mode, c10::SymInt hidden_size, c10::SymInt proj_size, int64_t num_layers, bool batch_first, bool bidirectional, at::Tensor & out);
+};
+
+}} // namespace at::_ops
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_atan_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_atan_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..add7838a1784a72c8f5527dad52300e48166f155
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_atan_native.h
@@ -0,0 +1,25 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API void _foreach_atan_out(at::TensorList self, at::TensorList out);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_atan_slow(at::TensorList self);
+TORCH_API void foreach_tensor_atan_slow_(at::TensorList self);
+TORCH_API ::std::vector<at::Tensor> foreach_tensor_atan_cuda(at::TensorList self);
+TORCH_API void foreach_tensor_atan_cuda_(at::TensorList self);
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_erfc_ops.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_erfc_ops.h
new file mode 100644
index 0000000000000000000000000000000000000000..5d8db13a8ccddbfe2cc7ea57945ca1f107d290fd
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_foreach_erfc_ops.h
@@ -0,0 +1,50 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API _foreach_erfc {
+  using schema = ::std::vector<at::Tensor> (at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_foreach_erfc")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_foreach_erfc(Tensor[] self) -> Tensor[]")
+  static ::std::vector<at::Tensor> call(at::TensorList self);
+  static ::std::vector<at::Tensor> redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_erfc_ {
+  using schema = void (at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_foreach_erfc_")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_foreach_erfc_(Tensor(a!)[] self) -> ()")
+  static void call(at::TensorList self);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self);
+};
+
+struct TORCH_API _foreach_erfc_out {
+  using schema = void (at::TensorList, at::TensorList);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::_foreach_erfc")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "_foreach_erfc.out(Tensor[] self, *, Tensor(a!)[] out) -> ()")
+  static void call(at::TensorList self, at::TensorList out);
+  static void redispatch(c10::DispatchKeySet dispatchKeySet, at::TensorList self, at::TensorList out);
+};
+
+}} // namespace at::_ops
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_int_mm_cuda_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_int_mm_cuda_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..2a312a781ba387eaba12f616e903957674981fd3
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_int_mm_cuda_dispatch.h
@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API at::Tensor _int_mm(const at::Tensor & self, const at::Tensor & mat2);
+TORCH_API at::Tensor & _int_mm_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & mat2);
+TORCH_API at::Tensor & _int_mm_outf(const at::Tensor & self, const at::Tensor & mat2, at::Tensor & out);
+
+} // namespace cuda
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_sparse_csr_prod.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_sparse_csr_prod.h
new file mode 100644
index 0000000000000000000000000000000000000000..b9290baae61fe8b96dffeeebb1e840c9d2949608
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_sparse_csr_prod.h
@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/_sparse_csr_prod_ops.h>
+
+namespace at {
+
+
+// aten::_sparse_csr_prod.dim_dtype(Tensor self, int[1] dim, bool keepdim=False, *, ScalarType? dtype=None) -> Tensor
+inline at::Tensor _sparse_csr_prod(const at::Tensor & self, at::IntArrayRef dim, bool keepdim=false, c10::optional<at::ScalarType> dtype=c10::nullopt) {
+    return at::_ops::_sparse_csr_prod_dim_dtype::call(self, dim, keepdim, dtype);
+}
+
+// aten::_sparse_csr_prod.dim_dtype_out(Tensor self, int[1] dim, bool keepdim=False, *, ScalarType? dtype=None, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _sparse_csr_prod_out(at::Tensor & out, const at::Tensor & self, at::IntArrayRef dim, bool keepdim=false, c10::optional<at::ScalarType> dtype=c10::nullopt) {
+    return at::_ops::_sparse_csr_prod_dim_dtype_out::call(self, dim, keepdim, dtype, out);
+}
+// aten::_sparse_csr_prod.dim_dtype_out(Tensor self, int[1] dim, bool keepdim=False, *, ScalarType? dtype=None, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & _sparse_csr_prod_outf(const at::Tensor & self, at::IntArrayRef dim, bool keepdim, c10::optional<at::ScalarType> dtype, at::Tensor & out) {
+    return at::_ops::_sparse_csr_prod_dim_dtype_out::call(self, dim, keepdim, dtype, out);
+}
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_sparse_log_softmax_backward_data_compositeexplicitautograd_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_sparse_log_softmax_backward_data_compositeexplicitautograd_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..01aab769ee9e990baa7720b039e7373a362177fa
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_sparse_log_softmax_backward_data_compositeexplicitautograd_dispatch.h
@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API at::Tensor & _sparse_log_softmax_backward_data_out(at::Tensor & out, const at::Tensor & grad_output, const at::Tensor & output, int64_t dim, const at::Tensor & self);
+TORCH_API at::Tensor & _sparse_log_softmax_backward_data_outf(const at::Tensor & grad_output, const at::Tensor & output, int64_t dim, const at::Tensor & self, at::Tensor & out);
+
+} // namespace compositeexplicitautograd
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_test_string_default.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_test_string_default.h
new file mode 100644
index 0000000000000000000000000000000000000000..3cca30bb2175812c33ee71ea4b22daad7bc2cf3c
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_test_string_default.h
@@ -0,0 +1,30 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/_test_string_default_ops.h>
+
+namespace at {
+
+
+// aten::_test_string_default(Tensor dummy, str a="\"'\\", str b='"\'\\') -> Tensor
+inline at::Tensor _test_string_default(const at::Tensor & dummy, c10::string_view a="\"'\\", c10::string_view b="\"'\\") {
+    return at::_ops::_test_string_default::call(dummy, a, b);
+}
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_test_string_default_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_test_string_default_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..2d64fcf47accf1ca5222c1156a039c953a4275ca
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_test_string_default_native.h
@@ -0,0 +1,21 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor _test_string_default(const at::Tensor & dummy, c10::string_view a="\"'\\", c10::string_view b="\"'\\");
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_to_sparse_csr_cpu_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_to_sparse_csr_cpu_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..d92b4dd06aa6537bd33b4b805f49ff11c15460da
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_to_sparse_csr_cpu_dispatch.h
@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor _to_sparse_csr(const at::Tensor & self, c10::optional<int64_t> dense_dim=c10::nullopt);
+
+} // namespace cpu
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_bicubic2d_aa_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_bicubic2d_aa_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..004d947aba7fbec382d9dcfe87f7ef3eee144e31
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_bicubic2d_aa_native.h
@@ -0,0 +1,27 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+#include <ATen/ops/_upsample_bicubic2d_aa_meta.h>
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor _upsample_bicubic2d_aa(const at::Tensor & input, at::OptionalIntArrayRef output_size, bool align_corners, c10::optional<at::ArrayRef<double>> scale_factors);
+struct TORCH_API structured__upsample_bicubic2d_aa_out_cpu : public at::meta::structured__upsample_bicubic2d_aa {
+void impl(const at::Tensor & self, at::ArrayRef<int64_t> output_size, bool align_corners, c10::optional<double> scales_h, c10::optional<double> scales_w, const at::Tensor & out);
+};
+struct TORCH_API structured__upsample_bicubic2d_aa_out_cuda : public at::meta::structured__upsample_bicubic2d_aa {
+void impl(const at::Tensor & self, at::ArrayRef<int64_t> output_size, bool align_corners, c10::optional<double> scales_h, c10::optional<double> scales_w, const at::Tensor & out);
+};
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact2d_backward_meta.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact2d_backward_meta.h
new file mode 100644
index 0000000000000000000000000000000000000000..cb821be2a9371eaa56b4d45531ad1925667f1e42
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact2d_backward_meta.h
@@ -0,0 +1,27 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeMetaFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/TensorIterator.h>
+#include <ATen/TensorMeta.h>
+#include <tuple>
+#include <vector>
+
+namespace at {
+namespace meta {
+
+struct TORCH_API structured__upsample_nearest_exact2d_backward : public at::impl::MetaBase {
+    
+    
+    void meta(const at::Tensor & grad_output, at::ArrayRef<int64_t> output_size, at::ArrayRef<int64_t> input_size, c10::optional<double> scales_h, c10::optional<double> scales_w);
+};
+
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact3d_cpu_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact3d_cpu_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..900275c9d12180cc79475a4c0129d58404e92ae8
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/_upsample_nearest_exact3d_cpu_dispatch.h
@@ -0,0 +1,28 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor _upsample_nearest_exact3d(const at::Tensor & self, at::IntArrayRef output_size, c10::optional<double> scales_d=c10::nullopt, c10::optional<double> scales_h=c10::nullopt, c10::optional<double> scales_w=c10::nullopt);
+TORCH_API at::Tensor _upsample_nearest_exact3d_symint(const at::Tensor & self, c10::SymIntArrayRef output_size, c10::optional<double> scales_d=c10::nullopt, c10::optional<double> scales_h=c10::nullopt, c10::optional<double> scales_w=c10::nullopt);
+TORCH_API at::Tensor & _upsample_nearest_exact3d_out(at::Tensor & out, const at::Tensor & self, at::IntArrayRef output_size, c10::optional<double> scales_d=c10::nullopt, c10::optional<double> scales_h=c10::nullopt, c10::optional<double> scales_w=c10::nullopt);
+TORCH_API at::Tensor & _upsample_nearest_exact3d_outf(const at::Tensor & self, at::IntArrayRef output_size, c10::optional<double> scales_d, c10::optional<double> scales_h, c10::optional<double> scales_w, at::Tensor & out);
+TORCH_API at::Tensor & _upsample_nearest_exact3d_symint_out(at::Tensor & out, const at::Tensor & self, c10::SymIntArrayRef output_size, c10::optional<double> scales_d=c10::nullopt, c10::optional<double> scales_h=c10::nullopt, c10::optional<double> scales_w=c10::nullopt);
+TORCH_API at::Tensor & _upsample_nearest_exact3d_symint_outf(const at::Tensor & self, c10::SymIntArrayRef output_size, c10::optional<double> scales_d, c10::optional<double> scales_h, c10::optional<double> scales_w, at::Tensor & out);
+
+} // namespace cpu
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/adaptive_avg_pool3d_backward_cpu_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/adaptive_avg_pool3d_backward_cpu_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..0bdc3b57dbdc573db11947d8c9292e26fd3e3dd7
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/adaptive_avg_pool3d_backward_cpu_dispatch.h
@@ -0,0 +1,24 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor & adaptive_avg_pool3d_backward_out(at::Tensor & grad_input, const at::Tensor & grad_output, const at::Tensor & self);
+TORCH_API at::Tensor & adaptive_avg_pool3d_backward_outf(const at::Tensor & grad_output, const at::Tensor & self, at::Tensor & grad_input);
+
+} // namespace cpu
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/adaptive_max_pool2d_backward_meta_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/adaptive_max_pool2d_backward_meta_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..dc0e2c2c95a5a685421a3fdbea4121a2075f118d
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/adaptive_max_pool2d_backward_meta_dispatch.h
@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace meta {
+
+TORCH_API at::Tensor adaptive_max_pool2d_backward(const at::Tensor & grad_output, const at::Tensor & self, const at::Tensor & indices);
+TORCH_API at::Tensor & adaptive_max_pool2d_backward_out(at::Tensor & grad_input, const at::Tensor & grad_output, const at::Tensor & self, const at::Tensor & indices);
+TORCH_API at::Tensor & adaptive_max_pool2d_backward_outf(const at::Tensor & grad_output, const at::Tensor & self, const at::Tensor & indices, at::Tensor & grad_input);
+
+} // namespace meta
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/add_compositeexplicitautograd_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/add_compositeexplicitautograd_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..935b184402a06f99844e2b8088dea574ec6ed0c1
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/add_compositeexplicitautograd_dispatch.h
@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautograd {
+
+TORCH_API at::Tensor add(const at::Tensor & self, const at::Scalar & other, const at::Scalar & alpha=1);
+TORCH_API at::Tensor & add_out(at::Tensor & out, const at::Tensor & self, const at::Scalar & other, const at::Scalar & alpha=1);
+TORCH_API at::Tensor & add_outf(const at::Tensor & self, const at::Scalar & other, const at::Scalar & alpha, at::Tensor & out);
+TORCH_API at::Tensor & add_(at::Tensor & self, const at::Scalar & other, const at::Scalar & alpha=1);
+
+} // namespace compositeexplicitautograd
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_scatter_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_scatter_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..b8231436159ad9ab4fd8420175adb12e67a27cf2
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/as_strided_scatter_native.h
@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor & as_strided_scatter_out_symint(const at::Tensor & self, const at::Tensor & src, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, c10::optional<c10::SymInt> storage_offset, at::Tensor & out);
+TORCH_API at::Tensor as_strided_scatter_symint(const at::Tensor & self, const at::Tensor & src, c10::SymIntArrayRef size, c10::SymIntArrayRef stride, c10::optional<c10::SymInt> storage_offset=c10::nullopt);
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/coalesce_ops.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/coalesce_ops.h
new file mode 100644
index 0000000000000000000000000000000000000000..626df101f74f73cdf17dd1eddeaf074490bdf1f2
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/coalesce_ops.h
@@ -0,0 +1,28 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API coalesce {
+  using schema = at::Tensor (const at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::coalesce")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "coalesce(Tensor(a) self) -> Tensor(a)")
+  static at::Tensor call(const at::Tensor & self);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self);
+};
+
+}} // namespace at::_ops
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/conj.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/conj.h
new file mode 100644
index 0000000000000000000000000000000000000000..6a80689ab9a78b36864e3aa42a4224e6810fcdf3
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/conj.h
@@ -0,0 +1,30 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/conj_ops.h>
+
+namespace at {
+
+
+// aten::conj(Tensor(a) self) -> Tensor(a)
+inline at::Tensor __dispatch_conj(const at::Tensor & self) {
+    return at::_ops::conj::call(self);
+}
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/conv_transpose2d.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/conv_transpose2d.h
new file mode 100644
index 0000000000000000000000000000000000000000..3c6818773b8f359d91894dbb3c8c4a207ed24ddb
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/conv_transpose2d.h
@@ -0,0 +1,47 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/conv_transpose2d_ops.h>
+
+namespace at {
+
+
+// aten::conv_transpose2d.input(Tensor input, Tensor weight, Tensor? bias=None, SymInt[2] stride=1, SymInt[2] padding=0, SymInt[2] output_padding=0, SymInt groups=1, SymInt[2] dilation=1) -> Tensor
+inline at::Tensor conv_transpose2d(const at::Tensor & input, const at::Tensor & weight, const c10::optional<at::Tensor> & bias={}, at::IntArrayRef stride=1, at::IntArrayRef padding=0, at::IntArrayRef output_padding=0, int64_t groups=1, at::IntArrayRef dilation=1) {
+    return at::_ops::conv_transpose2d_input::call(input, weight, bias, c10::fromIntArrayRefSlow(stride), c10::fromIntArrayRefSlow(padding), c10::fromIntArrayRefSlow(output_padding), groups, c10::fromIntArrayRefSlow(dilation));
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor conv_transpose2d(const at::Tensor & input, const at::Tensor & weight, const c10::optional<at::Tensor> & bias={}, at::IntArrayRef stride=1, at::IntArrayRef padding=0, at::IntArrayRef output_padding=0, int64_t groups=1, at::IntArrayRef dilation=1) {
+    return at::_ops::conv_transpose2d_input::call(input, weight, bias, c10::fromIntArrayRefSlow(stride), c10::fromIntArrayRefSlow(padding), c10::fromIntArrayRefSlow(output_padding), groups, c10::fromIntArrayRefSlow(dilation));
+  }
+}
+
+// aten::conv_transpose2d.input(Tensor input, Tensor weight, Tensor? bias=None, SymInt[2] stride=1, SymInt[2] padding=0, SymInt[2] output_padding=0, SymInt groups=1, SymInt[2] dilation=1) -> Tensor
+inline at::Tensor conv_transpose2d_symint(const at::Tensor & input, const at::Tensor & weight, const c10::optional<at::Tensor> & bias={}, c10::SymIntArrayRef stride=c10::SymInt(1), c10::SymIntArrayRef padding=c10::SymInt(0), c10::SymIntArrayRef output_padding=c10::SymInt(0), c10::SymInt groups=1, c10::SymIntArrayRef dilation=c10::SymInt(1)) {
+    return at::_ops::conv_transpose2d_input::call(input, weight, bias, stride, padding, output_padding, groups, dilation);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor conv_transpose2d(const at::Tensor & input, const at::Tensor & weight, const c10::optional<at::Tensor> & bias={}, c10::SymIntArrayRef stride=c10::SymInt(1), c10::SymIntArrayRef padding=c10::SymInt(0), c10::SymIntArrayRef output_padding=c10::SymInt(0), c10::SymInt groups=1, c10::SymIntArrayRef dilation=c10::SymInt(1)) {
+    return at::_ops::conv_transpose2d_input::call(input, weight, bias, stride, padding, output_padding, groups, dilation);
+  }
+}
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_batch_norm_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_batch_norm_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..e249876fc471620f11a210deaff4559ed74e44c4
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_batch_norm_native.h
@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API ::std::tuple<at::Tensor &,at::Tensor &,at::Tensor &,at::Tensor &> cudnn_batch_norm_out(const at::Tensor & input, const at::Tensor & weight, const c10::optional<at::Tensor> & bias, const c10::optional<at::Tensor> & running_mean, const c10::optional<at::Tensor> & running_var, bool training, double exponential_average_factor, double epsilon, at::Tensor & out0, at::Tensor & out1, at::Tensor & out2, at::Tensor & out3);
+TORCH_API ::std::tuple<at::Tensor,at::Tensor,at::Tensor,at::Tensor> cudnn_batch_norm(const at::Tensor & input, const at::Tensor & weight, const c10::optional<at::Tensor> & bias, const c10::optional<at::Tensor> & running_mean, const c10::optional<at::Tensor> & running_var, bool training, double exponential_average_factor, double epsilon);
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_convolution_transpose_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_convolution_transpose_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..55bfe1d981b1d276686d0d19be784437cc0c9458
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/cudnn_convolution_transpose_native.h
@@ -0,0 +1,22 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API at::Tensor & cudnn_convolution_transpose_out_symint(const at::Tensor & self, const at::Tensor & weight, c10::SymIntArrayRef padding, c10::SymIntArrayRef output_padding, c10::SymIntArrayRef stride, c10::SymIntArrayRef dilation, c10::SymInt groups, bool benchmark, bool deterministic, bool allow_tf32, at::Tensor & out);
+TORCH_API at::Tensor cudnn_convolution_transpose(const at::Tensor & self, const at::Tensor & weight, at::IntArrayRef padding, at::IntArrayRef output_padding, at::IntArrayRef stride, at::IntArrayRef dilation, int64_t groups, bool benchmark, bool deterministic, bool allow_tf32);
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/diagonal_backward_ops.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/diagonal_backward_ops.h
new file mode 100644
index 0000000000000000000000000000000000000000..869f138b1194711cbb7599d97d2900b5daaa7239
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/diagonal_backward_ops.h
@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API diagonal_backward {
+  using schema = at::Tensor (const at::Tensor &, c10::SymIntArrayRef, int64_t, int64_t, int64_t);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::diagonal_backward")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "diagonal_backward(Tensor grad_output, SymInt[] input_sizes, int offset, int dim1, int dim2) -> Tensor")
+  static at::Tensor call(const at::Tensor & grad_output, c10::SymIntArrayRef input_sizes, int64_t offset, int64_t dim1, int64_t dim2);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad_output, c10::SymIntArrayRef input_sizes, int64_t offset, int64_t dim1, int64_t dim2);
+};
+
+struct TORCH_API diagonal_backward_out {
+  using schema = at::Tensor & (const at::Tensor &, c10::SymIntArrayRef, int64_t, int64_t, int64_t, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::diagonal_backward")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "diagonal_backward.out(Tensor grad_output, SymInt[] input_sizes, int offset, int dim1, int dim2, *, Tensor(a!) out) -> Tensor(a!)")
+  static at::Tensor & call(const at::Tensor & grad_output, c10::SymIntArrayRef input_sizes, int64_t offset, int64_t dim1, int64_t dim2, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & grad_output, c10::SymIntArrayRef input_sizes, int64_t offset, int64_t dim1, int64_t dim2, at::Tensor & out);
+};
+
+}} // namespace at::_ops
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/erfinv_meta.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/erfinv_meta.h
new file mode 100644
index 0000000000000000000000000000000000000000..764e4120340797b9a9ab49409fa273c4796cb26c
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/erfinv_meta.h
@@ -0,0 +1,27 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeMetaFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/TensorIterator.h>
+#include <ATen/TensorMeta.h>
+#include <tuple>
+#include <vector>
+
+namespace at {
+namespace meta {
+
+struct TORCH_API structured_erfinv : public TensorIteratorBase {
+    
+    
+    void meta(const at::Tensor & self);
+};
+
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/igammac_cuda_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/igammac_cuda_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..802b83632fb36510e1859ba333f67df84360298c
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/igammac_cuda_dispatch.h
@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API at::Tensor igammac(const at::Tensor & self, const at::Tensor & other);
+TORCH_API at::Tensor & igammac_out(at::Tensor & out, const at::Tensor & self, const at::Tensor & other);
+TORCH_API at::Tensor & igammac_outf(const at::Tensor & self, const at::Tensor & other, at::Tensor & out);
+TORCH_API at::Tensor & igammac_(at::Tensor & self, const at::Tensor & other);
+
+} // namespace cuda
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/logit_backward_cuda_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/logit_backward_cuda_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..b8c5e70a7d5413e344355d2cf85f22335477d2e1
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/logit_backward_cuda_dispatch.h
@@ -0,0 +1,25 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cuda {
+
+TORCH_API at::Tensor logit_backward(const at::Tensor & grad_output, const at::Tensor & self, c10::optional<double> eps=c10::nullopt);
+TORCH_API at::Tensor & logit_backward_out(at::Tensor & grad_input, const at::Tensor & grad_output, const at::Tensor & self, c10::optional<double> eps=c10::nullopt);
+TORCH_API at::Tensor & logit_backward_outf(const at::Tensor & grad_output, const at::Tensor & self, c10::optional<double> eps, at::Tensor & grad_input);
+
+} // namespace cuda
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/mH.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/mH.h
new file mode 100644
index 0000000000000000000000000000000000000000..a5f142b402135e9b421e210c4713ff3253f7ef70
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/mH.h
@@ -0,0 +1,26 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/mH_ops.h>
+
+namespace at {
+
+
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/margin_ranking_loss_compositeimplicitautograd_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/margin_ranking_loss_compositeimplicitautograd_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..ea56b707a0ac0e9398f61c2b7ed8d815afa88f64
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/margin_ranking_loss_compositeimplicitautograd_dispatch.h
@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeimplicitautograd {
+
+TORCH_API at::Tensor margin_ranking_loss(const at::Tensor & input1, const at::Tensor & input2, const at::Tensor & target, double margin=0.0, int64_t reduction=at::Reduction::Mean);
+
+} // namespace compositeimplicitautograd
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/max_pool2d_with_indices_backward_compositeexplicitautogradnonfunctional_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/max_pool2d_with_indices_backward_compositeexplicitautogradnonfunctional_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..3ff969183a4b7be6dac728dbc965f73e89b12012
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/max_pool2d_with_indices_backward_compositeexplicitautogradnonfunctional_dispatch.h
@@ -0,0 +1,23 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace compositeexplicitautogradnonfunctional {
+
+TORCH_API at::Tensor max_pool2d_with_indices_backward(const at::Tensor & grad_output, const at::Tensor & self, at::IntArrayRef kernel_size, at::IntArrayRef stride, at::IntArrayRef padding, at::IntArrayRef dilation, bool ceil_mode, const at::Tensor & indices);
+
+} // namespace compositeexplicitautogradnonfunctional
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/neg_cpu_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/neg_cpu_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..5218e0f2a31f6bb178f98d074cb6f42795056b77
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/neg_cpu_dispatch.h
@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor neg(const at::Tensor & self);
+TORCH_API at::Tensor & neg_out(at::Tensor & out, const at::Tensor & self);
+TORCH_API at::Tensor & neg_outf(const at::Tensor & self, at::Tensor & out);
+TORCH_API at::Tensor & neg_(at::Tensor & self);
+
+} // namespace cpu
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/pairwise_distance.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/pairwise_distance.h
new file mode 100644
index 0000000000000000000000000000000000000000..b48a625a1973bdfdb0f8f51fa60c55f73acbd27d
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/pairwise_distance.h
@@ -0,0 +1,30 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/pairwise_distance_ops.h>
+
+namespace at {
+
+
+// aten::pairwise_distance(Tensor x1, Tensor x2, float p=2, float eps=1e-06, bool keepdim=False) -> Tensor
+inline at::Tensor pairwise_distance(const at::Tensor & x1, const at::Tensor & x2, double p=2, double eps=1e-06, bool keepdim=false) {
+    return at::_ops::pairwise_distance::call(x1, x2, p, eps, keepdim);
+}
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/randn.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/randn.h
new file mode 100644
index 0000000000000000000000000000000000000000..920e942d328d03fc4bde24f308c28ed745565754
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/randn.h
@@ -0,0 +1,377 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Function.h
+
+#include <ATen/Context.h>
+#include <ATen/DeviceGuard.h>
+#include <ATen/TensorUtils.h>
+#include <ATen/TracerMode.h>
+#include <ATen/core/Generator.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+
+
+
+#include <ATen/ops/randn_ops.h>
+
+namespace at {
+
+
+// aten::randn(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::randn::call(c10::fromIntArrayRefSlow(size), optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::randn::call(c10::fromIntArrayRefSlow(size), optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn::call(c10::fromIntArrayRefSlow(size), dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn::call(c10::fromIntArrayRefSlow(size), dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::randn::call(size, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, at::TensorOptions options={}) {
+    return at::_ops::randn::call(size, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn(SymInt[] size, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn::call(size, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn::call(size, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.generator(SymInt[] size, *, Generator? generator, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, at::TensorOptions options={}) {
+    return at::_ops::randn_generator::call(c10::fromIntArrayRefSlow(size), generator, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, at::TensorOptions options={}) {
+    return at::_ops::randn_generator::call(c10::fromIntArrayRefSlow(size), generator, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn.generator(SymInt[] size, *, Generator? generator, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator::call(c10::fromIntArrayRefSlow(size), generator, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator::call(c10::fromIntArrayRefSlow(size), generator, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.generator(SymInt[] size, *, Generator? generator, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, at::TensorOptions options={}) {
+    return at::_ops::randn_generator::call(size, generator, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, at::TensorOptions options={}) {
+    return at::_ops::randn_generator::call(size, generator, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn.generator(SymInt[] size, *, Generator? generator, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator::call(size, generator, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator::call(size, generator, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.names(SymInt[] size, *, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_names::call(c10::fromIntArrayRefSlow(size), names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_names::call(c10::fromIntArrayRefSlow(size), names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn.names(SymInt[] size, *, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_names::call(c10::fromIntArrayRefSlow(size), names, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_names::call(c10::fromIntArrayRefSlow(size), names, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.names(SymInt[] size, *, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_names::call(size, names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_names::call(size, names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn.names(SymInt[] size, *, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_names::call(size, names, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_names::call(size, names, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.generator_with_names(SymInt[] size, *, Generator? generator, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_generator_with_names::call(c10::fromIntArrayRefSlow(size), generator, names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_generator_with_names::call(c10::fromIntArrayRefSlow(size), generator, names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn.generator_with_names(SymInt[] size, *, Generator? generator, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator_with_names::call(c10::fromIntArrayRefSlow(size), generator, names, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor randn(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator_with_names::call(c10::fromIntArrayRefSlow(size), generator, names, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.generator_with_names(SymInt[] size, *, Generator? generator, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_generator_with_names::call(size, generator, names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::TensorOptions options={}) {
+    return at::_ops::randn_generator_with_names::call(size, generator, names, optTypeMetaToScalarType(options.dtype_opt()), options.layout_opt(), options.device_opt(), options.pinned_memory_opt());
+  }
+}
+
+// aten::randn.generator_with_names(SymInt[] size, *, Generator? generator, Dimname[]? names, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None) -> Tensor
+inline at::Tensor randn_symint(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator_with_names::call(size, generator, names, dtype, layout, device, pin_memory);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor randn(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory) {
+    return at::_ops::randn_generator_with_names::call(size, generator, names, dtype, layout, device, pin_memory);
+  }
+}
+
+// aten::randn.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size) {
+    return at::_ops::randn_out::call(c10::fromIntArrayRefSlow(size), out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size) {
+    return at::_ops::randn_out::call(c10::fromIntArrayRefSlow(size), out);
+  }
+}
+
+// aten::randn.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_outf(at::IntArrayRef size, at::Tensor & out) {
+    return at::_ops::randn_out::call(c10::fromIntArrayRefSlow(size), out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_outf(at::IntArrayRef size, at::Tensor & out) {
+    return at::_ops::randn_out::call(c10::fromIntArrayRefSlow(size), out);
+  }
+}
+
+// aten::randn.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_out(at::Tensor & out, c10::SymIntArrayRef size) {
+    return at::_ops::randn_out::call(size, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_out(at::Tensor & out, c10::SymIntArrayRef size) {
+    return at::_ops::randn_out::call(size, out);
+  }
+}
+
+// aten::randn.out(SymInt[] size, *, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_outf(c10::SymIntArrayRef size, at::Tensor & out) {
+    return at::_ops::randn_out::call(size, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_outf(c10::SymIntArrayRef size, at::Tensor & out) {
+    return at::_ops::randn_out::call(size, out);
+  }
+}
+
+// aten::randn.generator_out(SymInt[] size, *, Generator? generator, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size, c10::optional<at::Generator> generator) {
+    return at::_ops::randn_generator_out::call(c10::fromIntArrayRefSlow(size), generator, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size, c10::optional<at::Generator> generator) {
+    return at::_ops::randn_generator_out::call(c10::fromIntArrayRefSlow(size), generator, out);
+  }
+}
+
+// aten::randn.generator_out(SymInt[] size, *, Generator? generator, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_outf(at::IntArrayRef size, c10::optional<at::Generator> generator, at::Tensor & out) {
+    return at::_ops::randn_generator_out::call(c10::fromIntArrayRefSlow(size), generator, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_outf(at::IntArrayRef size, c10::optional<at::Generator> generator, at::Tensor & out) {
+    return at::_ops::randn_generator_out::call(c10::fromIntArrayRefSlow(size), generator, out);
+  }
+}
+
+// aten::randn.generator_out(SymInt[] size, *, Generator? generator, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_out(at::Tensor & out, c10::SymIntArrayRef size, c10::optional<at::Generator> generator) {
+    return at::_ops::randn_generator_out::call(size, generator, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_out(at::Tensor & out, c10::SymIntArrayRef size, c10::optional<at::Generator> generator) {
+    return at::_ops::randn_generator_out::call(size, generator, out);
+  }
+}
+
+// aten::randn.generator_out(SymInt[] size, *, Generator? generator, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_outf(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, at::Tensor & out) {
+    return at::_ops::randn_generator_out::call(size, generator, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_outf(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, at::Tensor & out) {
+    return at::_ops::randn_generator_out::call(size, generator, out);
+  }
+}
+
+// aten::randn.names_out(SymInt[] size, *, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_names_out::call(c10::fromIntArrayRefSlow(size), names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_names_out::call(c10::fromIntArrayRefSlow(size), names, out);
+  }
+}
+
+// aten::randn.names_out(SymInt[] size, *, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_outf(at::IntArrayRef size, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_names_out::call(c10::fromIntArrayRefSlow(size), names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_outf(at::IntArrayRef size, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_names_out::call(c10::fromIntArrayRefSlow(size), names, out);
+  }
+}
+
+// aten::randn.names_out(SymInt[] size, *, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_out(at::Tensor & out, c10::SymIntArrayRef size, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_names_out::call(size, names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_out(at::Tensor & out, c10::SymIntArrayRef size, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_names_out::call(size, names, out);
+  }
+}
+
+// aten::randn.names_out(SymInt[] size, *, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_outf(c10::SymIntArrayRef size, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_names_out::call(size, names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_outf(c10::SymIntArrayRef size, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_names_out::call(size, names, out);
+  }
+}
+
+// aten::randn.generator_with_names_out(SymInt[] size, *, Generator? generator, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_generator_with_names_out::call(c10::fromIntArrayRefSlow(size), generator, names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_out(at::Tensor & out, at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_generator_with_names_out::call(c10::fromIntArrayRefSlow(size), generator, names, out);
+  }
+}
+
+// aten::randn.generator_with_names_out(SymInt[] size, *, Generator? generator, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_outf(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_generator_with_names_out::call(c10::fromIntArrayRefSlow(size), generator, names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, int64_t>::value>>
+  at::Tensor & randn_outf(at::IntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_generator_with_names_out::call(c10::fromIntArrayRefSlow(size), generator, names, out);
+  }
+}
+
+// aten::randn.generator_with_names_out(SymInt[] size, *, Generator? generator, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_out(at::Tensor & out, c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_generator_with_names_out::call(size, generator, names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_out(at::Tensor & out, c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names) {
+    return at::_ops::randn_generator_with_names_out::call(size, generator, names, out);
+  }
+}
+
+// aten::randn.generator_with_names_out(SymInt[] size, *, Generator? generator, Dimname[]? names, Tensor(a!) out) -> Tensor(a!)
+inline at::Tensor & randn_symint_outf(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_generator_with_names_out::call(size, generator, names, out);
+}
+namespace symint {
+  template <typename T, typename = std::enable_if_t<std::is_same<T, c10::SymInt>::value>>
+  at::Tensor & randn_outf(c10::SymIntArrayRef size, c10::optional<at::Generator> generator, c10::optional<at::DimnameList> names, at::Tensor & out) {
+    return at::_ops::randn_generator_with_names_out::call(size, generator, names, out);
+  }
+}
+
+}
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/scatter_reduce_meta_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/scatter_reduce_meta_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..4dba6685f9cd4f1a64bd32688981acd777631822
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/scatter_reduce_meta_dispatch.h
@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace meta {
+
+TORCH_API at::Tensor scatter_reduce(const at::Tensor & self, int64_t dim, const at::Tensor & index, const at::Tensor & src, c10::string_view reduce, bool include_self=true);
+TORCH_API at::Tensor & scatter_reduce_out(at::Tensor & out, const at::Tensor & self, int64_t dim, const at::Tensor & index, const at::Tensor & src, c10::string_view reduce, bool include_self=true);
+TORCH_API at::Tensor & scatter_reduce_outf(const at::Tensor & self, int64_t dim, const at::Tensor & index, const at::Tensor & src, c10::string_view reduce, bool include_self, at::Tensor & out);
+TORCH_API at::Tensor & scatter_reduce_(at::Tensor & self, int64_t dim, const at::Tensor & index, const at::Tensor & src, c10::string_view reduce, bool include_self=true);
+
+} // namespace meta
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/sigmoid_cpu_dispatch.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/sigmoid_cpu_dispatch.h
new file mode 100644
index 0000000000000000000000000000000000000000..2e38e35a09c06cb5161e3bc04f22902d1eaa024c
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/sigmoid_cpu_dispatch.h
@@ -0,0 +1,26 @@
+#pragma once
+// @generated by torchgen/gen.py from DispatchKeyFunction.h
+
+// NB: The implementing C++ file is RegisterDispatchKey.cpp
+
+// The only #includes we need are for custom classes that have defaults in the C++ API
+#include <c10/core/MemoryFormat.h>
+#include <c10/core/Scalar.h>
+#include <ATen/core/Reduction.h>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+
+namespace cpu {
+
+TORCH_API at::Tensor sigmoid(const at::Tensor & self);
+TORCH_API at::Tensor & sigmoid_out(at::Tensor & out, const at::Tensor & self);
+TORCH_API at::Tensor & sigmoid_outf(const at::Tensor & self, at::Tensor & out);
+TORCH_API at::Tensor & sigmoid_(at::Tensor & self);
+
+} // namespace cpu
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/sparse_dim_native.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/sparse_dim_native.h
new file mode 100644
index 0000000000000000000000000000000000000000..b20cba1153d65864a13c5d8fedfa5bd58b1a9697
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/sparse_dim_native.h
@@ -0,0 +1,23 @@
+#pragma once
+
+// @generated by torchgen/gen.py from NativeFunction.h
+
+#include <c10/core/Scalar.h>
+#include <c10/core/Storage.h>
+#include <c10/core/TensorOptions.h>
+#include <c10/util/Deprecated.h>
+#include <c10/util/Optional.h>
+#include <c10/core/QScheme.h>
+#include <ATen/core/Reduction.h>
+#include <ATen/core/Tensor.h>
+#include <tuple>
+#include <vector>
+
+
+namespace at {
+namespace native {
+TORCH_API int64_t sparse_dim_strided(const at::Tensor & self);
+TORCH_API int64_t sparse_dim_sparse(const at::Tensor & self);
+TORCH_API int64_t sparse_dim_sparse_csr(const at::Tensor & self);
+} // namespace native
+} // namespace at
diff --git a/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/zeros_like_ops.h b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/zeros_like_ops.h
new file mode 100644
index 0000000000000000000000000000000000000000..a5bed006e1a6a00259e6cc9ee1b5a9ff7be215fb
--- /dev/null
+++ b/videollama2/lib/python3.10/site-packages/torch/include/ATen/ops/zeros_like_ops.h
@@ -0,0 +1,39 @@
+#pragma once
+
+// @generated by torchgen/gen.py from Operator.h
+
+#include <tuple>
+#include <vector>
+
+// Forward declarations of any types needed in the operator signatures.
+// We can't directly include these classes because it will cause circular include dependencies.
+// This file is included by TensorBody.h, which defines the Tensor class.
+#include <ATen/core/ATen_fwd.h>
+
+namespace at {
+namespace _ops {
+
+
+struct TORCH_API zeros_like {
+  using schema = at::Tensor (const at::Tensor &, c10::optional<at::ScalarType>, c10::optional<at::Layout>, c10::optional<at::Device>, c10::optional<bool>, c10::optional<at::MemoryFormat>);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::zeros_like")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "zeros_like(Tensor self, *, ScalarType? dtype=None, Layout? layout=None, Device? device=None, bool? pin_memory=None, MemoryFormat? memory_format=None) -> Tensor")
+  static at::Tensor call(const at::Tensor & self, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory, c10::optional<at::MemoryFormat> memory_format);
+  static at::Tensor redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, c10::optional<at::ScalarType> dtype, c10::optional<at::Layout> layout, c10::optional<at::Device> device, c10::optional<bool> pin_memory, c10::optional<at::MemoryFormat> memory_format);
+};
+
+struct TORCH_API zeros_like_out {
+  using schema = at::Tensor & (const at::Tensor &, c10::optional<at::MemoryFormat>, at::Tensor &);
+  using ptr_schema = schema*;
+  // See Note [static constexpr char* members for windows NVCC]
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(name, "aten::zeros_like")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(overload_name, "out")
+  STATIC_CONSTEXPR_STR_INL_EXCEPT_WIN_CUDA(schema_str, "zeros_like.out(Tensor self, *, MemoryFormat? memory_format=None, Tensor(a!) out) -> Tensor(a!)")
+  static at::Tensor & call(const at::Tensor & self, c10::optional<at::MemoryFormat> memory_format, at::Tensor & out);
+  static at::Tensor & redispatch(c10::DispatchKeySet dispatchKeySet, const at::Tensor & self, c10::optional<at::MemoryFormat> memory_format, at::Tensor & out);
+};
+
+}} // namespace at::_ops
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__init__.py b/vllm/lib/python3.10/site-packages/py/_code/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..f15acf85132113f090b1a53cdd0bac39af924d4d
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/__init__.py
@@ -0,0 +1 @@
+""" python inspection/code generation API """
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/__init__.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7a15f9a6347fa4333b34ab595940b92028723a57
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/__init__.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_assertionnew.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_assertionnew.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..3b1bf62d7651212bb1179014c6e6094150372d3c
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_assertionnew.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_assertionold.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_assertionold.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9a449610c688515865009ea715f4c80255c8d1af
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_assertionold.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_py2traceback.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_py2traceback.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..d88f702f8c38859ff0c7a9ba8e6f75aa812ab858
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/_py2traceback.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/assertion.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/assertion.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8921fb0e533a0d449ba1ab301249b934f8d83a70
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/assertion.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/code.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/code.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..97009feca4b70268e991c13a550f3f15b9b4cef9
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/code.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/__pycache__/source.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/source.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..cde8e50121f9e14d8c44dcebefd6458a3a526be4
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_code/__pycache__/source.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_code/_assertionnew.py b/vllm/lib/python3.10/site-packages/py/_code/_assertionnew.py
new file mode 100644
index 0000000000000000000000000000000000000000..d03f29d870814283d3f2d537d91b8a527821ee1f
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/_assertionnew.py
@@ -0,0 +1,322 @@
+"""
+Find intermediate evalutation results in assert statements through builtin AST.
+This should replace _assertionold.py eventually.
+"""
+
+import sys
+import ast
+
+import py
+from py._code.assertion import _format_explanation, BuiltinAssertionError
+
+
+def _is_ast_expr(node):
+    return isinstance(node, ast.expr)
+def _is_ast_stmt(node):
+    return isinstance(node, ast.stmt)
+
+
+class Failure(Exception):
+    """Error found while interpreting AST."""
+
+    def __init__(self, explanation=""):
+        self.cause = sys.exc_info()
+        self.explanation = explanation
+
+
+def interpret(source, frame, should_fail=False):
+    mod = ast.parse(source)
+    visitor = DebugInterpreter(frame)
+    try:
+        visitor.visit(mod)
+    except Failure:
+        failure = sys.exc_info()[1]
+        return getfailure(failure)
+    if should_fail:
+        return ("(assertion failed, but when it was re-run for "
+                "printing intermediate values, it did not fail.  Suggestions: "
+                "compute assert expression before the assert or use --no-assert)")
+
+def run(offending_line, frame=None):
+    if frame is None:
+        frame = py.code.Frame(sys._getframe(1))
+    return interpret(offending_line, frame)
+
+def getfailure(failure):
+    explanation = _format_explanation(failure.explanation)
+    value = failure.cause[1]
+    if str(value):
+        lines = explanation.splitlines()
+        if not lines:
+            lines.append("")
+        lines[0] += " << %s" % (value,)
+        explanation = "\n".join(lines)
+    text = "%s: %s" % (failure.cause[0].__name__, explanation)
+    if text.startswith("AssertionError: assert "):
+        text = text[16:]
+    return text
+
+
+operator_map = {
+    ast.BitOr : "|",
+    ast.BitXor : "^",
+    ast.BitAnd : "&",
+    ast.LShift : "<<",
+    ast.RShift : ">>",
+    ast.Add : "+",
+    ast.Sub : "-",
+    ast.Mult : "*",
+    ast.Div : "/",
+    ast.FloorDiv : "//",
+    ast.Mod : "%",
+    ast.Eq : "==",
+    ast.NotEq : "!=",
+    ast.Lt : "<",
+    ast.LtE : "<=",
+    ast.Gt : ">",
+    ast.GtE : ">=",
+    ast.Pow : "**",
+    ast.Is : "is",
+    ast.IsNot : "is not",
+    ast.In : "in",
+    ast.NotIn : "not in"
+}
+
+unary_map = {
+    ast.Not : "not %s",
+    ast.Invert : "~%s",
+    ast.USub : "-%s",
+    ast.UAdd : "+%s"
+}
+
+
+class DebugInterpreter(ast.NodeVisitor):
+    """Interpret AST nodes to gleam useful debugging information. """
+
+    def __init__(self, frame):
+        self.frame = frame
+
+    def generic_visit(self, node):
+        # Fallback when we don't have a special implementation.
+        if _is_ast_expr(node):
+            mod = ast.Expression(node)
+            co = self._compile(mod)
+            try:
+                result = self.frame.eval(co)
+            except Exception:
+                raise Failure()
+            explanation = self.frame.repr(result)
+            return explanation, result
+        elif _is_ast_stmt(node):
+            mod = ast.Module([node])
+            co = self._compile(mod, "exec")
+            try:
+                self.frame.exec_(co)
+            except Exception:
+                raise Failure()
+            return None, None
+        else:
+            raise AssertionError("can't handle %s" %(node,))
+
+    def _compile(self, source, mode="eval"):
+        return compile(source, "<assertion interpretation>", mode)
+
+    def visit_Expr(self, expr):
+        return self.visit(expr.value)
+
+    def visit_Module(self, mod):
+        for stmt in mod.body:
+            self.visit(stmt)
+
+    def visit_Name(self, name):
+        explanation, result = self.generic_visit(name)
+        # See if the name is local.
+        source = "%r in locals() is not globals()" % (name.id,)
+        co = self._compile(source)
+        try:
+            local = self.frame.eval(co)
+        except Exception:
+            # have to assume it isn't
+            local = False
+        if not local:
+            return name.id, result
+        return explanation, result
+
+    def visit_Compare(self, comp):
+        left = comp.left
+        left_explanation, left_result = self.visit(left)
+        for op, next_op in zip(comp.ops, comp.comparators):
+            next_explanation, next_result = self.visit(next_op)
+            op_symbol = operator_map[op.__class__]
+            explanation = "%s %s %s" % (left_explanation, op_symbol,
+                                        next_explanation)
+            source = "__exprinfo_left %s __exprinfo_right" % (op_symbol,)
+            co = self._compile(source)
+            try:
+                result = self.frame.eval(co, __exprinfo_left=left_result,
+                                         __exprinfo_right=next_result)
+            except Exception:
+                raise Failure(explanation)
+            try:
+                if not result:
+                    break
+            except KeyboardInterrupt:
+                raise
+            except:
+                break
+            left_explanation, left_result = next_explanation, next_result
+
+        rcomp = py.code._reprcompare
+        if rcomp:
+            res = rcomp(op_symbol, left_result, next_result)
+            if res:
+                explanation = res
+        return explanation, result
+
+    def visit_BoolOp(self, boolop):
+        is_or = isinstance(boolop.op, ast.Or)
+        explanations = []
+        for operand in boolop.values:
+            explanation, result = self.visit(operand)
+            explanations.append(explanation)
+            if result == is_or:
+                break
+        name = is_or and " or " or " and "
+        explanation = "(" + name.join(explanations) + ")"
+        return explanation, result
+
+    def visit_UnaryOp(self, unary):
+        pattern = unary_map[unary.op.__class__]
+        operand_explanation, operand_result = self.visit(unary.operand)
+        explanation = pattern % (operand_explanation,)
+        co = self._compile(pattern % ("__exprinfo_expr",))
+        try:
+            result = self.frame.eval(co, __exprinfo_expr=operand_result)
+        except Exception:
+            raise Failure(explanation)
+        return explanation, result
+
+    def visit_BinOp(self, binop):
+        left_explanation, left_result = self.visit(binop.left)
+        right_explanation, right_result = self.visit(binop.right)
+        symbol = operator_map[binop.op.__class__]
+        explanation = "(%s %s %s)" % (left_explanation, symbol,
+                                      right_explanation)
+        source = "__exprinfo_left %s __exprinfo_right" % (symbol,)
+        co = self._compile(source)
+        try:
+            result = self.frame.eval(co, __exprinfo_left=left_result,
+                                     __exprinfo_right=right_result)
+        except Exception:
+            raise Failure(explanation)
+        return explanation, result
+
+    def visit_Call(self, call):
+        func_explanation, func = self.visit(call.func)
+        arg_explanations = []
+        ns = {"__exprinfo_func" : func}
+        arguments = []
+        for arg in call.args:
+            arg_explanation, arg_result = self.visit(arg)
+            arg_name = "__exprinfo_%s" % (len(ns),)
+            ns[arg_name] = arg_result
+            arguments.append(arg_name)
+            arg_explanations.append(arg_explanation)
+        for keyword in call.keywords:
+            arg_explanation, arg_result = self.visit(keyword.value)
+            arg_name = "__exprinfo_%s" % (len(ns),)
+            ns[arg_name] = arg_result
+            keyword_source = "%s=%%s" % (keyword.arg)
+            arguments.append(keyword_source % (arg_name,))
+            arg_explanations.append(keyword_source % (arg_explanation,))
+        if call.starargs:
+            arg_explanation, arg_result = self.visit(call.starargs)
+            arg_name = "__exprinfo_star"
+            ns[arg_name] = arg_result
+            arguments.append("*%s" % (arg_name,))
+            arg_explanations.append("*%s" % (arg_explanation,))
+        if call.kwargs:
+            arg_explanation, arg_result = self.visit(call.kwargs)
+            arg_name = "__exprinfo_kwds"
+            ns[arg_name] = arg_result
+            arguments.append("**%s" % (arg_name,))
+            arg_explanations.append("**%s" % (arg_explanation,))
+        args_explained = ", ".join(arg_explanations)
+        explanation = "%s(%s)" % (func_explanation, args_explained)
+        args = ", ".join(arguments)
+        source = "__exprinfo_func(%s)" % (args,)
+        co = self._compile(source)
+        try:
+            result = self.frame.eval(co, **ns)
+        except Exception:
+            raise Failure(explanation)
+        pattern = "%s\n{%s = %s\n}"
+        rep = self.frame.repr(result)
+        explanation = pattern % (rep, rep, explanation)
+        return explanation, result
+
+    def _is_builtin_name(self, name):
+        pattern = "%r not in globals() and %r not in locals()"
+        source = pattern % (name.id, name.id)
+        co = self._compile(source)
+        try:
+            return self.frame.eval(co)
+        except Exception:
+            return False
+
+    def visit_Attribute(self, attr):
+        if not isinstance(attr.ctx, ast.Load):
+            return self.generic_visit(attr)
+        source_explanation, source_result = self.visit(attr.value)
+        explanation = "%s.%s" % (source_explanation, attr.attr)
+        source = "__exprinfo_expr.%s" % (attr.attr,)
+        co = self._compile(source)
+        try:
+            result = self.frame.eval(co, __exprinfo_expr=source_result)
+        except Exception:
+            raise Failure(explanation)
+        explanation = "%s\n{%s = %s.%s\n}" % (self.frame.repr(result),
+                                              self.frame.repr(result),
+                                              source_explanation, attr.attr)
+        # Check if the attr is from an instance.
+        source = "%r in getattr(__exprinfo_expr, '__dict__', {})"
+        source = source % (attr.attr,)
+        co = self._compile(source)
+        try:
+            from_instance = self.frame.eval(co, __exprinfo_expr=source_result)
+        except Exception:
+            from_instance = True
+        if from_instance:
+            rep = self.frame.repr(result)
+            pattern = "%s\n{%s = %s\n}"
+            explanation = pattern % (rep, rep, explanation)
+        return explanation, result
+
+    def visit_Assert(self, assrt):
+        test_explanation, test_result = self.visit(assrt.test)
+        if test_explanation.startswith("False\n{False =") and \
+                test_explanation.endswith("\n"):
+            test_explanation = test_explanation[15:-2]
+        explanation = "assert %s" % (test_explanation,)
+        if not test_result:
+            try:
+                raise BuiltinAssertionError
+            except Exception:
+                raise Failure(explanation)
+        return explanation, test_result
+
+    def visit_Assign(self, assign):
+        value_explanation, value_result = self.visit(assign.value)
+        explanation = "... = %s" % (value_explanation,)
+        name = ast.Name("__exprinfo_expr", ast.Load(),
+                        lineno=assign.value.lineno,
+                        col_offset=assign.value.col_offset)
+        new_assign = ast.Assign(assign.targets, name, lineno=assign.lineno,
+                                col_offset=assign.col_offset)
+        mod = ast.Module([new_assign])
+        co = self._compile(mod, "exec")
+        try:
+            self.frame.exec_(co, __exprinfo_expr=value_result)
+        except Exception:
+            raise Failure(explanation)
+        return explanation, value_result
diff --git a/vllm/lib/python3.10/site-packages/py/_code/_assertionold.py b/vllm/lib/python3.10/site-packages/py/_code/_assertionold.py
new file mode 100644
index 0000000000000000000000000000000000000000..1bb70a875d059d664c6518701963864afad86593
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/_assertionold.py
@@ -0,0 +1,556 @@
+import py
+import sys, inspect
+from compiler import parse, ast, pycodegen
+from py._code.assertion import BuiltinAssertionError, _format_explanation
+import types
+
+passthroughex = py.builtin._sysex
+
+class Failure:
+    def __init__(self, node):
+        self.exc, self.value, self.tb = sys.exc_info()
+        self.node = node
+
+class View(object):
+    """View base class.
+
+    If C is a subclass of View, then C(x) creates a proxy object around
+    the object x.  The actual class of the proxy is not C in general,
+    but a *subclass* of C determined by the rules below.  To avoid confusion
+    we call view class the class of the proxy (a subclass of C, so of View)
+    and object class the class of x.
+
+    Attributes and methods not found in the proxy are automatically read on x.
+    Other operations like setting attributes are performed on the proxy, as
+    determined by its view class.  The object x is available from the proxy
+    as its __obj__ attribute.
+
+    The view class selection is determined by the __view__ tuples and the
+    optional __viewkey__ method.  By default, the selected view class is the
+    most specific subclass of C whose __view__ mentions the class of x.
+    If no such subclass is found, the search proceeds with the parent
+    object classes.  For example, C(True) will first look for a subclass
+    of C with __view__ = (..., bool, ...) and only if it doesn't find any
+    look for one with __view__ = (..., int, ...), and then ..., object,...
+    If everything fails the class C itself is considered to be the default.
+
+    Alternatively, the view class selection can be driven by another aspect
+    of the object x, instead of the class of x, by overriding __viewkey__.
+    See last example at the end of this module.
+    """
+
+    _viewcache = {}
+    __view__ = ()
+
+    def __new__(rootclass, obj, *args, **kwds):
+        self = object.__new__(rootclass)
+        self.__obj__ = obj
+        self.__rootclass__ = rootclass
+        key = self.__viewkey__()
+        try:
+            self.__class__ = self._viewcache[key]
+        except KeyError:
+            self.__class__ = self._selectsubclass(key)
+        return self
+
+    def __getattr__(self, attr):
+        # attributes not found in the normal hierarchy rooted on View
+        # are looked up in the object's real class
+        return getattr(self.__obj__, attr)
+
+    def __viewkey__(self):
+        return self.__obj__.__class__
+
+    def __matchkey__(self, key, subclasses):
+        if inspect.isclass(key):
+            keys = inspect.getmro(key)
+        else:
+            keys = [key]
+        for key in keys:
+            result = [C for C in subclasses if key in C.__view__]
+            if result:
+                return result
+        return []
+
+    def _selectsubclass(self, key):
+        subclasses = list(enumsubclasses(self.__rootclass__))
+        for C in subclasses:
+            if not isinstance(C.__view__, tuple):
+                C.__view__ = (C.__view__,)
+        choices = self.__matchkey__(key, subclasses)
+        if not choices:
+            return self.__rootclass__
+        elif len(choices) == 1:
+            return choices[0]
+        else:
+            # combine the multiple choices
+            return type('?', tuple(choices), {})
+
+    def __repr__(self):
+        return '%s(%r)' % (self.__rootclass__.__name__, self.__obj__)
+
+
+def enumsubclasses(cls):
+    for subcls in cls.__subclasses__():
+        for subsubclass in enumsubclasses(subcls):
+            yield subsubclass
+    yield cls
+
+
+class Interpretable(View):
+    """A parse tree node with a few extra methods."""
+    explanation = None
+
+    def is_builtin(self, frame):
+        return False
+
+    def eval(self, frame):
+        # fall-back for unknown expression nodes
+        try:
+            expr = ast.Expression(self.__obj__)
+            expr.filename = '<eval>'
+            self.__obj__.filename = '<eval>'
+            co = pycodegen.ExpressionCodeGenerator(expr).getCode()
+            result = frame.eval(co)
+        except passthroughex:
+            raise
+        except:
+            raise Failure(self)
+        self.result = result
+        self.explanation = self.explanation or frame.repr(self.result)
+
+    def run(self, frame):
+        # fall-back for unknown statement nodes
+        try:
+            expr = ast.Module(None, ast.Stmt([self.__obj__]))
+            expr.filename = '<run>'
+            co = pycodegen.ModuleCodeGenerator(expr).getCode()
+            frame.exec_(co)
+        except passthroughex:
+            raise
+        except:
+            raise Failure(self)
+
+    def nice_explanation(self):
+        return _format_explanation(self.explanation)
+
+
+class Name(Interpretable):
+    __view__ = ast.Name
+
+    def is_local(self, frame):
+        source = '%r in locals() is not globals()' % self.name
+        try:
+            return frame.is_true(frame.eval(source))
+        except passthroughex:
+            raise
+        except:
+            return False
+
+    def is_global(self, frame):
+        source = '%r in globals()' % self.name
+        try:
+            return frame.is_true(frame.eval(source))
+        except passthroughex:
+            raise
+        except:
+            return False
+
+    def is_builtin(self, frame):
+        source = '%r not in locals() and %r not in globals()' % (
+            self.name, self.name)
+        try:
+            return frame.is_true(frame.eval(source))
+        except passthroughex:
+            raise
+        except:
+            return False
+
+    def eval(self, frame):
+        super(Name, self).eval(frame)
+        if not self.is_local(frame):
+            self.explanation = self.name
+
+class Compare(Interpretable):
+    __view__ = ast.Compare
+
+    def eval(self, frame):
+        expr = Interpretable(self.expr)
+        expr.eval(frame)
+        for operation, expr2 in self.ops:
+            if hasattr(self, 'result'):
+                # shortcutting in chained expressions
+                if not frame.is_true(self.result):
+                    break
+            expr2 = Interpretable(expr2)
+            expr2.eval(frame)
+            self.explanation = "%s %s %s" % (
+                expr.explanation, operation, expr2.explanation)
+            source = "__exprinfo_left %s __exprinfo_right" % operation
+            try:
+                self.result = frame.eval(source,
+                                         __exprinfo_left=expr.result,
+                                         __exprinfo_right=expr2.result)
+            except passthroughex:
+                raise
+            except:
+                raise Failure(self)
+            expr = expr2
+
+class And(Interpretable):
+    __view__ = ast.And
+
+    def eval(self, frame):
+        explanations = []
+        for expr in self.nodes:
+            expr = Interpretable(expr)
+            expr.eval(frame)
+            explanations.append(expr.explanation)
+            self.result = expr.result
+            if not frame.is_true(expr.result):
+                break
+        self.explanation = '(' + ' and '.join(explanations) + ')'
+
+class Or(Interpretable):
+    __view__ = ast.Or
+
+    def eval(self, frame):
+        explanations = []
+        for expr in self.nodes:
+            expr = Interpretable(expr)
+            expr.eval(frame)
+            explanations.append(expr.explanation)
+            self.result = expr.result
+            if frame.is_true(expr.result):
+                break
+        self.explanation = '(' + ' or '.join(explanations) + ')'
+
+
+# == Unary operations ==
+keepalive = []
+for astclass, astpattern in {
+    ast.Not    : 'not __exprinfo_expr',
+    ast.Invert : '(~__exprinfo_expr)',
+    }.items():
+
+    class UnaryArith(Interpretable):
+        __view__ = astclass
+
+        def eval(self, frame, astpattern=astpattern):
+            expr = Interpretable(self.expr)
+            expr.eval(frame)
+            self.explanation = astpattern.replace('__exprinfo_expr',
+                                                  expr.explanation)
+            try:
+                self.result = frame.eval(astpattern,
+                                         __exprinfo_expr=expr.result)
+            except passthroughex:
+                raise
+            except:
+                raise Failure(self)
+
+    keepalive.append(UnaryArith)
+
+# == Binary operations ==
+for astclass, astpattern in {
+    ast.Add    : '(__exprinfo_left + __exprinfo_right)',
+    ast.Sub    : '(__exprinfo_left - __exprinfo_right)',
+    ast.Mul    : '(__exprinfo_left * __exprinfo_right)',
+    ast.Div    : '(__exprinfo_left / __exprinfo_right)',
+    ast.Mod    : '(__exprinfo_left % __exprinfo_right)',
+    ast.Power  : '(__exprinfo_left ** __exprinfo_right)',
+    }.items():
+
+    class BinaryArith(Interpretable):
+        __view__ = astclass
+
+        def eval(self, frame, astpattern=astpattern):
+            left = Interpretable(self.left)
+            left.eval(frame)
+            right = Interpretable(self.right)
+            right.eval(frame)
+            self.explanation = (astpattern
+                                .replace('__exprinfo_left',  left .explanation)
+                                .replace('__exprinfo_right', right.explanation))
+            try:
+                self.result = frame.eval(astpattern,
+                                         __exprinfo_left=left.result,
+                                         __exprinfo_right=right.result)
+            except passthroughex:
+                raise
+            except:
+                raise Failure(self)
+
+    keepalive.append(BinaryArith)
+
+
+class CallFunc(Interpretable):
+    __view__ = ast.CallFunc
+
+    def is_bool(self, frame):
+        source = 'isinstance(__exprinfo_value, bool)'
+        try:
+            return frame.is_true(frame.eval(source,
+                                            __exprinfo_value=self.result))
+        except passthroughex:
+            raise
+        except:
+            return False
+
+    def eval(self, frame):
+        node = Interpretable(self.node)
+        node.eval(frame)
+        explanations = []
+        vars = {'__exprinfo_fn': node.result}
+        source = '__exprinfo_fn('
+        for a in self.args:
+            if isinstance(a, ast.Keyword):
+                keyword = a.name
+                a = a.expr
+            else:
+                keyword = None
+            a = Interpretable(a)
+            a.eval(frame)
+            argname = '__exprinfo_%d' % len(vars)
+            vars[argname] = a.result
+            if keyword is None:
+                source += argname + ','
+                explanations.append(a.explanation)
+            else:
+                source += '%s=%s,' % (keyword, argname)
+                explanations.append('%s=%s' % (keyword, a.explanation))
+        if self.star_args:
+            star_args = Interpretable(self.star_args)
+            star_args.eval(frame)
+            argname = '__exprinfo_star'
+            vars[argname] = star_args.result
+            source += '*' + argname + ','
+            explanations.append('*' + star_args.explanation)
+        if self.dstar_args:
+            dstar_args = Interpretable(self.dstar_args)
+            dstar_args.eval(frame)
+            argname = '__exprinfo_kwds'
+            vars[argname] = dstar_args.result
+            source += '**' + argname + ','
+            explanations.append('**' + dstar_args.explanation)
+        self.explanation = "%s(%s)" % (
+            node.explanation, ', '.join(explanations))
+        if source.endswith(','):
+            source = source[:-1]
+        source += ')'
+        try:
+            self.result = frame.eval(source, **vars)
+        except passthroughex:
+            raise
+        except:
+            raise Failure(self)
+        if not node.is_builtin(frame) or not self.is_bool(frame):
+            r = frame.repr(self.result)
+            self.explanation = '%s\n{%s = %s\n}' % (r, r, self.explanation)
+
+class Getattr(Interpretable):
+    __view__ = ast.Getattr
+
+    def eval(self, frame):
+        expr = Interpretable(self.expr)
+        expr.eval(frame)
+        source = '__exprinfo_expr.%s' % self.attrname
+        try:
+            self.result = frame.eval(source, __exprinfo_expr=expr.result)
+        except passthroughex:
+            raise
+        except:
+            raise Failure(self)
+        self.explanation = '%s.%s' % (expr.explanation, self.attrname)
+        # if the attribute comes from the instance, its value is interesting
+        source = ('hasattr(__exprinfo_expr, "__dict__") and '
+                  '%r in __exprinfo_expr.__dict__' % self.attrname)
+        try:
+            from_instance = frame.is_true(
+                frame.eval(source, __exprinfo_expr=expr.result))
+        except passthroughex:
+            raise
+        except:
+            from_instance = True
+        if from_instance:
+            r = frame.repr(self.result)
+            self.explanation = '%s\n{%s = %s\n}' % (r, r, self.explanation)
+
+# == Re-interpretation of full statements ==
+
+class Assert(Interpretable):
+    __view__ = ast.Assert
+
+    def run(self, frame):
+        test = Interpretable(self.test)
+        test.eval(frame)
+        # simplify 'assert False where False = ...'
+        if (test.explanation.startswith('False\n{False = ') and
+            test.explanation.endswith('\n}')):
+            test.explanation = test.explanation[15:-2]
+        # print the result as  'assert <explanation>'
+        self.result = test.result
+        self.explanation = 'assert ' + test.explanation
+        if not frame.is_true(test.result):
+            try:
+                raise BuiltinAssertionError
+            except passthroughex:
+                raise
+            except:
+                raise Failure(self)
+
+class Assign(Interpretable):
+    __view__ = ast.Assign
+
+    def run(self, frame):
+        expr = Interpretable(self.expr)
+        expr.eval(frame)
+        self.result = expr.result
+        self.explanation = '... = ' + expr.explanation
+        # fall-back-run the rest of the assignment
+        ass = ast.Assign(self.nodes, ast.Name('__exprinfo_expr'))
+        mod = ast.Module(None, ast.Stmt([ass]))
+        mod.filename = '<run>'
+        co = pycodegen.ModuleCodeGenerator(mod).getCode()
+        try:
+            frame.exec_(co, __exprinfo_expr=expr.result)
+        except passthroughex:
+            raise
+        except:
+            raise Failure(self)
+
+class Discard(Interpretable):
+    __view__ = ast.Discard
+
+    def run(self, frame):
+        expr = Interpretable(self.expr)
+        expr.eval(frame)
+        self.result = expr.result
+        self.explanation = expr.explanation
+
+class Stmt(Interpretable):
+    __view__ = ast.Stmt
+
+    def run(self, frame):
+        for stmt in self.nodes:
+            stmt = Interpretable(stmt)
+            stmt.run(frame)
+
+
+def report_failure(e):
+    explanation = e.node.nice_explanation()
+    if explanation:
+        explanation = ", in: " + explanation
+    else:
+        explanation = ""
+    sys.stdout.write("%s: %s%s\n" % (e.exc.__name__, e.value, explanation))
+
+def check(s, frame=None):
+    if frame is None:
+        frame = sys._getframe(1)
+        frame = py.code.Frame(frame)
+    expr = parse(s, 'eval')
+    assert isinstance(expr, ast.Expression)
+    node = Interpretable(expr.node)
+    try:
+        node.eval(frame)
+    except passthroughex:
+        raise
+    except Failure:
+        e = sys.exc_info()[1]
+        report_failure(e)
+    else:
+        if not frame.is_true(node.result):
+            sys.stderr.write("assertion failed: %s\n" % node.nice_explanation())
+
+
+###########################################################
+# API / Entry points
+# #########################################################
+
+def interpret(source, frame, should_fail=False):
+    module = Interpretable(parse(source, 'exec').node)
+    #print "got module", module
+    if isinstance(frame, types.FrameType):
+        frame = py.code.Frame(frame)
+    try:
+        module.run(frame)
+    except Failure:
+        e = sys.exc_info()[1]
+        return getfailure(e)
+    except passthroughex:
+        raise
+    except:
+        import traceback
+        traceback.print_exc()
+    if should_fail:
+        return ("(assertion failed, but when it was re-run for "
+                "printing intermediate values, it did not fail.  Suggestions: "
+                "compute assert expression before the assert or use --nomagic)")
+    else:
+        return None
+
+def getmsg(excinfo):
+    if isinstance(excinfo, tuple):
+        excinfo = py.code.ExceptionInfo(excinfo)
+    #frame, line = gettbline(tb)
+    #frame = py.code.Frame(frame)
+    #return interpret(line, frame)
+
+    tb = excinfo.traceback[-1]
+    source = str(tb.statement).strip()
+    x = interpret(source, tb.frame, should_fail=True)
+    if not isinstance(x, str):
+        raise TypeError("interpret returned non-string %r" % (x,))
+    return x
+
+def getfailure(e):
+    explanation = e.node.nice_explanation()
+    if str(e.value):
+        lines = explanation.split('\n')
+        lines[0] += "  << %s" % (e.value,)
+        explanation = '\n'.join(lines)
+    text = "%s: %s" % (e.exc.__name__, explanation)
+    if text.startswith('AssertionError: assert '):
+        text = text[16:]
+    return text
+
+def run(s, frame=None):
+    if frame is None:
+        frame = sys._getframe(1)
+        frame = py.code.Frame(frame)
+    module = Interpretable(parse(s, 'exec').node)
+    try:
+        module.run(frame)
+    except Failure:
+        e = sys.exc_info()[1]
+        report_failure(e)
+
+
+if __name__ == '__main__':
+    # example:
+    def f():
+        return 5
+    def g():
+        return 3
+    def h(x):
+        return 'never'
+    check("f() * g() == 5")
+    check("not f()")
+    check("not (f() and g() or 0)")
+    check("f() == g()")
+    i = 4
+    check("i == f()")
+    check("len(f()) == 0")
+    check("isinstance(2+3+4, float)")
+
+    run("x = i")
+    check("x == 5")
+
+    run("assert not f(), 'oops'")
+    run("a, b, c = 1, 2")
+    run("a, b, c = f()")
+
+    check("max([f(),g()]) == 4")
+    check("'hello'[g()] == 'h'")
+    run("'guk%d' % h(f())")
diff --git a/vllm/lib/python3.10/site-packages/py/_code/_py2traceback.py b/vllm/lib/python3.10/site-packages/py/_code/_py2traceback.py
new file mode 100644
index 0000000000000000000000000000000000000000..d65e27cb73077bbd33bc0fad0d20a89e443bef9e
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/_py2traceback.py
@@ -0,0 +1,79 @@
+# copied from python-2.7.3's traceback.py
+# CHANGES:
+# - some_str is replaced, trying to create unicode strings
+#
+import types
+
+def format_exception_only(etype, value):
+    """Format the exception part of a traceback.
+
+    The arguments are the exception type and value such as given by
+    sys.last_type and sys.last_value. The return value is a list of
+    strings, each ending in a newline.
+
+    Normally, the list contains a single string; however, for
+    SyntaxError exceptions, it contains several lines that (when
+    printed) display detailed information about where the syntax
+    error occurred.
+
+    The message indicating which exception occurred is always the last
+    string in the list.
+
+    """
+
+    # An instance should not have a meaningful value parameter, but
+    # sometimes does, particularly for string exceptions, such as
+    # >>> raise string1, string2  # deprecated
+    #
+    # Clear these out first because issubtype(string1, SyntaxError)
+    # would throw another exception and mask the original problem.
+    if (isinstance(etype, BaseException) or
+        isinstance(etype, types.InstanceType) or
+        etype is None or type(etype) is str):
+        return [_format_final_exc_line(etype, value)]
+
+    stype = etype.__name__
+
+    if not issubclass(etype, SyntaxError):
+        return [_format_final_exc_line(stype, value)]
+
+    # It was a syntax error; show exactly where the problem was found.
+    lines = []
+    try:
+        msg, (filename, lineno, offset, badline) = value.args
+    except Exception:
+        pass
+    else:
+        filename = filename or "<string>"
+        lines.append('  File "%s", line %d\n' % (filename, lineno))
+        if badline is not None:
+            lines.append('    %s\n' % badline.strip())
+            if offset is not None:
+                caretspace = badline.rstrip('\n')[:offset].lstrip()
+                # non-space whitespace (likes tabs) must be kept for alignment
+                caretspace = ((c.isspace() and c or ' ') for c in caretspace)
+                # only three spaces to account for offset1 == pos 0
+                lines.append('   %s^\n' % ''.join(caretspace))
+        value = msg
+
+    lines.append(_format_final_exc_line(stype, value))
+    return lines
+
+def _format_final_exc_line(etype, value):
+    """Return a list of a single line -- normal case for format_exception_only"""
+    valuestr = _some_str(value)
+    if value is None or not valuestr:
+        line = "%s\n" % etype
+    else:
+        line = "%s: %s\n" % (etype, valuestr)
+    return line
+
+def _some_str(value):
+    try:
+        return unicode(value)
+    except Exception:
+        try:
+            return str(value)
+        except Exception:
+            pass
+    return '<unprintable %s object>' % type(value).__name__
diff --git a/vllm/lib/python3.10/site-packages/py/_code/assertion.py b/vllm/lib/python3.10/site-packages/py/_code/assertion.py
new file mode 100644
index 0000000000000000000000000000000000000000..ff1643799c9e015a8e351b89958ae3eb8111d668
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/assertion.py
@@ -0,0 +1,90 @@
+import sys
+import py
+
+BuiltinAssertionError = py.builtin.builtins.AssertionError
+
+_reprcompare = None # if set, will be called by assert reinterp for comparison ops
+
+def _format_explanation(explanation):
+    """This formats an explanation
+
+    Normally all embedded newlines are escaped, however there are
+    three exceptions: \n{, \n} and \n~.  The first two are intended
+    cover nested explanations, see function and attribute explanations
+    for examples (.visit_Call(), visit_Attribute()).  The last one is
+    for when one explanation needs to span multiple lines, e.g. when
+    displaying diffs.
+    """
+    raw_lines = (explanation or '').split('\n')
+    # escape newlines not followed by {, } and ~
+    lines = [raw_lines[0]]
+    for l in raw_lines[1:]:
+        if l.startswith('{') or l.startswith('}') or l.startswith('~'):
+            lines.append(l)
+        else:
+            lines[-1] += '\\n' + l
+
+    result = lines[:1]
+    stack = [0]
+    stackcnt = [0]
+    for line in lines[1:]:
+        if line.startswith('{'):
+            if stackcnt[-1]:
+                s = 'and   '
+            else:
+                s = 'where '
+            stack.append(len(result))
+            stackcnt[-1] += 1
+            stackcnt.append(0)
+            result.append(' +' + '  '*(len(stack)-1) + s + line[1:])
+        elif line.startswith('}'):
+            assert line.startswith('}')
+            stack.pop()
+            stackcnt.pop()
+            result[stack[-1]] += line[1:]
+        else:
+            assert line.startswith('~')
+            result.append('  '*len(stack) + line[1:])
+    assert len(stack) == 1
+    return '\n'.join(result)
+
+
+class AssertionError(BuiltinAssertionError):
+    def __init__(self, *args):
+        BuiltinAssertionError.__init__(self, *args)
+        if args:
+            try:
+                self.msg = str(args[0])
+            except py.builtin._sysex:
+                raise
+            except:
+                self.msg = "<[broken __repr__] %s at %0xd>" %(
+                    args[0].__class__, id(args[0]))
+        else:
+            f = py.code.Frame(sys._getframe(1))
+            try:
+                source = f.code.fullsource
+                if source is not None:
+                    try:
+                        source = source.getstatement(f.lineno, assertion=True)
+                    except IndexError:
+                        source = None
+                    else:
+                        source = str(source.deindent()).strip()
+            except py.error.ENOENT:
+                source = None
+                # this can also occur during reinterpretation, when the
+                # co_filename is set to "<run>".
+            if source:
+                self.msg = reinterpret(source, f, should_fail=True)
+            else:
+                self.msg = "<could not determine information>"
+            if not self.args:
+                self.args = (self.msg,)
+
+if sys.version_info > (3, 0):
+    AssertionError.__module__ = "builtins"
+    reinterpret_old = "old reinterpretation not available for py3"
+else:
+    from py._code._assertionold import interpret as reinterpret_old
+from py._code._assertionnew import interpret as reinterpret
diff --git a/vllm/lib/python3.10/site-packages/py/_code/code.py b/vllm/lib/python3.10/site-packages/py/_code/code.py
new file mode 100644
index 0000000000000000000000000000000000000000..dad796283fe64c89f5f99f2bc72adc4a7657d238
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/code.py
@@ -0,0 +1,796 @@
+import py
+import sys
+from inspect import CO_VARARGS, CO_VARKEYWORDS, isclass
+
+builtin_repr = repr
+
+reprlib = py.builtin._tryimport('repr', 'reprlib')
+
+if sys.version_info[0] >= 3:
+    from traceback import format_exception_only
+else:
+    from py._code._py2traceback import format_exception_only
+
+import traceback
+
+
+class Code(object):
+    """ wrapper around Python code objects """
+    def __init__(self, rawcode):
+        if not hasattr(rawcode, "co_filename"):
+            rawcode = py.code.getrawcode(rawcode)
+        try:
+            self.filename = rawcode.co_filename
+            self.firstlineno = rawcode.co_firstlineno - 1
+            self.name = rawcode.co_name
+        except AttributeError:
+            raise TypeError("not a code object: %r" % (rawcode,))
+        self.raw = rawcode
+
+    def __eq__(self, other):
+        return self.raw == other.raw
+
+    def __ne__(self, other):
+        return not self == other
+
+    @property
+    def path(self):
+        """ return a path object pointing to source code (note that it
+        might not point to an actually existing file). """
+        p = py.path.local(self.raw.co_filename)
+        # maybe don't try this checking
+        if not p.check():
+            # XXX maybe try harder like the weird logic
+            # in the standard lib [linecache.updatecache] does?
+            p = self.raw.co_filename
+        return p
+
+    @property
+    def fullsource(self):
+        """ return a py.code.Source object for the full source file of the code
+        """
+        from py._code import source
+        full, _ = source.findsource(self.raw)
+        return full
+
+    def source(self):
+        """ return a py.code.Source object for the code object's source only
+        """
+        # return source only for that part of code
+        return py.code.Source(self.raw)
+
+    def getargs(self, var=False):
+        """ return a tuple with the argument names for the code object
+
+            if 'var' is set True also return the names of the variable and
+            keyword arguments when present
+        """
+        # handfull shortcut for getting args
+        raw = self.raw
+        argcount = raw.co_argcount
+        if var:
+            argcount += raw.co_flags & CO_VARARGS
+            argcount += raw.co_flags & CO_VARKEYWORDS
+        return raw.co_varnames[:argcount]
+
+class Frame(object):
+    """Wrapper around a Python frame holding f_locals and f_globals
+    in which expressions can be evaluated."""
+
+    def __init__(self, frame):
+        self.lineno = frame.f_lineno - 1
+        self.f_globals = frame.f_globals
+        self.f_locals = frame.f_locals
+        self.raw = frame
+        self.code = py.code.Code(frame.f_code)
+
+    @property
+    def statement(self):
+        """ statement this frame is at """
+        if self.code.fullsource is None:
+            return py.code.Source("")
+        return self.code.fullsource.getstatement(self.lineno)
+
+    def eval(self, code, **vars):
+        """ evaluate 'code' in the frame
+
+            'vars' are optional additional local variables
+
+            returns the result of the evaluation
+        """
+        f_locals = self.f_locals.copy()
+        f_locals.update(vars)
+        return eval(code, self.f_globals, f_locals)
+
+    def exec_(self, code, **vars):
+        """ exec 'code' in the frame
+
+            'vars' are optiona; additional local variables
+        """
+        f_locals = self.f_locals.copy()
+        f_locals.update(vars)
+        py.builtin.exec_(code, self.f_globals, f_locals)
+
+    def repr(self, object):
+        """ return a 'safe' (non-recursive, one-line) string repr for 'object'
+        """
+        return py.io.saferepr(object)
+
+    def is_true(self, object):
+        return object
+
+    def getargs(self, var=False):
+        """ return a list of tuples (name, value) for all arguments
+
+            if 'var' is set True also include the variable and keyword
+            arguments when present
+        """
+        retval = []
+        for arg in self.code.getargs(var):
+            try:
+                retval.append((arg, self.f_locals[arg]))
+            except KeyError:
+                pass     # this can occur when using Psyco
+        return retval
+
+
+class TracebackEntry(object):
+    """ a single entry in a traceback """
+
+    _repr_style = None
+    exprinfo = None
+
+    def __init__(self, rawentry):
+        self._rawentry = rawentry
+        self.lineno = rawentry.tb_lineno - 1
+
+    def set_repr_style(self, mode):
+        assert mode in ("short", "long")
+        self._repr_style = mode
+
+    @property
+    def frame(self):
+        return py.code.Frame(self._rawentry.tb_frame)
+
+    @property
+    def relline(self):
+        return self.lineno - self.frame.code.firstlineno
+
+    def __repr__(self):
+        return "<TracebackEntry %s:%d>" % (self.frame.code.path, self.lineno+1)
+
+    @property
+    def statement(self):
+        """ py.code.Source object for the current statement """
+        source = self.frame.code.fullsource
+        return source.getstatement(self.lineno)
+
+    @property
+    def path(self):
+        """ path to the source code """
+        return self.frame.code.path
+
+    def getlocals(self):
+        return self.frame.f_locals
+    locals = property(getlocals, None, None, "locals of underlaying frame")
+
+    def reinterpret(self):
+        """Reinterpret the failing statement and returns a detailed information
+           about what operations are performed."""
+        if self.exprinfo is None:
+            source = str(self.statement).strip()
+            x = py.code._reinterpret(source, self.frame, should_fail=True)
+            if not isinstance(x, str):
+                raise TypeError("interpret returned non-string %r" % (x,))
+            self.exprinfo = x
+        return self.exprinfo
+
+    def getfirstlinesource(self):
+        # on Jython this firstlineno can be -1 apparently
+        return max(self.frame.code.firstlineno, 0)
+
+    def getsource(self, astcache=None):
+        """ return failing source code. """
+        # we use the passed in astcache to not reparse asttrees
+        # within exception info printing
+        from py._code.source import getstatementrange_ast
+        source = self.frame.code.fullsource
+        if source is None:
+            return None
+        key = astnode = None
+        if astcache is not None:
+            key = self.frame.code.path
+            if key is not None:
+                astnode = astcache.get(key, None)
+        start = self.getfirstlinesource()
+        try:
+            astnode, _, end = getstatementrange_ast(self.lineno, source,
+                                                    astnode=astnode)
+        except SyntaxError:
+            end = self.lineno + 1
+        else:
+            if key is not None:
+                astcache[key] = astnode
+        return source[start:end]
+
+    source = property(getsource)
+
+    def ishidden(self):
+        """ return True if the current frame has a var __tracebackhide__
+            resolving to True
+
+            mostly for internal use
+        """
+        try:
+            return self.frame.f_locals['__tracebackhide__']
+        except KeyError:
+            try:
+                return self.frame.f_globals['__tracebackhide__']
+            except KeyError:
+                return False
+
+    def __str__(self):
+        try:
+            fn = str(self.path)
+        except py.error.Error:
+            fn = '???'
+        name = self.frame.code.name
+        try:
+            line = str(self.statement).lstrip()
+        except KeyboardInterrupt:
+            raise
+        except:
+            line = "???"
+        return "  File %r:%d in %s\n  %s\n" % (fn, self.lineno+1, name, line)
+
+    def name(self):
+        return self.frame.code.raw.co_name
+    name = property(name, None, None, "co_name of underlaying code")
+
+
+class Traceback(list):
+    """ Traceback objects encapsulate and offer higher level
+        access to Traceback entries.
+    """
+    Entry = TracebackEntry
+
+    def __init__(self, tb):
+        """ initialize from given python traceback object. """
+        if hasattr(tb, 'tb_next'):
+            def f(cur):
+                while cur is not None:
+                    yield self.Entry(cur)
+                    cur = cur.tb_next
+            list.__init__(self, f(tb))
+        else:
+            list.__init__(self, tb)
+
+    def cut(self, path=None, lineno=None, firstlineno=None, excludepath=None):
+        """ return a Traceback instance wrapping part of this Traceback
+
+            by provding any combination of path, lineno and firstlineno, the
+            first frame to start the to-be-returned traceback is determined
+
+            this allows cutting the first part of a Traceback instance e.g.
+            for formatting reasons (removing some uninteresting bits that deal
+            with handling of the exception/traceback)
+        """
+        for x in self:
+            code = x.frame.code
+            codepath = code.path
+            if ((path is None or codepath == path) and
+                (excludepath is None or not hasattr(codepath, 'relto') or
+                 not codepath.relto(excludepath)) and
+                (lineno is None or x.lineno == lineno) and
+                (firstlineno is None or x.frame.code.firstlineno == firstlineno)):
+                return Traceback(x._rawentry)
+        return self
+
+    def __getitem__(self, key):
+        val = super(Traceback, self).__getitem__(key)
+        if isinstance(key, type(slice(0))):
+            val = self.__class__(val)
+        return val
+
+    def filter(self, fn=lambda x: not x.ishidden()):
+        """ return a Traceback instance with certain items removed
+
+            fn is a function that gets a single argument, a TracebackItem
+            instance, and should return True when the item should be added
+            to the Traceback, False when not
+
+            by default this removes all the TracebackItems which are hidden
+            (see ishidden() above)
+        """
+        return Traceback(filter(fn, self))
+
+    def getcrashentry(self):
+        """ return last non-hidden traceback entry that lead
+        to the exception of a traceback.
+        """
+        for i in range(-1, -len(self)-1, -1):
+            entry = self[i]
+            if not entry.ishidden():
+                return entry
+        return self[-1]
+
+    def recursionindex(self):
+        """ return the index of the frame/TracebackItem where recursion
+            originates if appropriate, None if no recursion occurred
+        """
+        cache = {}
+        for i, entry in enumerate(self):
+            # id for the code.raw is needed to work around
+            # the strange metaprogramming in the decorator lib from pypi
+            # which generates code objects that have hash/value equality
+            #XXX needs a test
+            key = entry.frame.code.path, id(entry.frame.code.raw), entry.lineno
+            #print "checking for recursion at", key
+            l = cache.setdefault(key, [])
+            if l:
+                f = entry.frame
+                loc = f.f_locals
+                for otherloc in l:
+                    if f.is_true(f.eval(co_equal,
+                        __recursioncache_locals_1=loc,
+                        __recursioncache_locals_2=otherloc)):
+                        return i
+            l.append(entry.frame.f_locals)
+        return None
+
+co_equal = compile('__recursioncache_locals_1 == __recursioncache_locals_2',
+                   '?', 'eval')
+
+class ExceptionInfo(object):
+    """ wraps sys.exc_info() objects and offers
+        help for navigating the traceback.
+    """
+    _striptext = ''
+    def __init__(self, tup=None, exprinfo=None):
+        if tup is None:
+            tup = sys.exc_info()
+            if exprinfo is None and isinstance(tup[1], AssertionError):
+                exprinfo = getattr(tup[1], 'msg', None)
+                if exprinfo is None:
+                    exprinfo = str(tup[1])
+                if exprinfo and exprinfo.startswith('assert '):
+                    self._striptext = 'AssertionError: '
+        self._excinfo = tup
+        #: the exception class
+        self.type = tup[0]
+        #: the exception instance
+        self.value = tup[1]
+        #: the exception raw traceback
+        self.tb = tup[2]
+        #: the exception type name
+        self.typename = self.type.__name__
+        #: the exception traceback (py.code.Traceback instance)
+        self.traceback = py.code.Traceback(self.tb)
+
+    def __repr__(self):
+        return "<ExceptionInfo %s tblen=%d>" % (
+            self.typename, len(self.traceback))
+
+    def exconly(self, tryshort=False):
+        """ return the exception as a string
+
+            when 'tryshort' resolves to True, and the exception is a
+            py.code._AssertionError, only the actual exception part of
+            the exception representation is returned (so 'AssertionError: ' is
+            removed from the beginning)
+        """
+        lines = format_exception_only(self.type, self.value)
+        text = ''.join(lines)
+        text = text.rstrip()
+        if tryshort:
+            if text.startswith(self._striptext):
+                text = text[len(self._striptext):]
+        return text
+
+    def errisinstance(self, exc):
+        """ return True if the exception is an instance of exc """
+        return isinstance(self.value, exc)
+
+    def _getreprcrash(self):
+        exconly = self.exconly(tryshort=True)
+        entry = self.traceback.getcrashentry()
+        path, lineno = entry.frame.code.raw.co_filename, entry.lineno
+        return ReprFileLocation(path, lineno+1, exconly)
+
+    def getrepr(self, showlocals=False, style="long",
+                abspath=False, tbfilter=True, funcargs=False):
+        """ return str()able representation of this exception info.
+            showlocals: show locals per traceback entry
+            style: long|short|no|native traceback style
+            tbfilter: hide entries (where __tracebackhide__ is true)
+
+            in case of style==native, tbfilter and showlocals is ignored.
+        """
+        if style == 'native':
+            return ReprExceptionInfo(ReprTracebackNative(
+                traceback.format_exception(
+                    self.type,
+                    self.value,
+                    self.traceback[0]._rawentry,
+                )), self._getreprcrash())
+
+        fmt = FormattedExcinfo(
+            showlocals=showlocals, style=style,
+            abspath=abspath, tbfilter=tbfilter, funcargs=funcargs)
+        return fmt.repr_excinfo(self)
+
+    def __str__(self):
+        entry = self.traceback[-1]
+        loc = ReprFileLocation(entry.path, entry.lineno + 1, self.exconly())
+        return str(loc)
+
+    def __unicode__(self):
+        entry = self.traceback[-1]
+        loc = ReprFileLocation(entry.path, entry.lineno + 1, self.exconly())
+        return loc.__unicode__()
+
+
+class FormattedExcinfo(object):
+    """ presenting information about failing Functions and Generators. """
+    # for traceback entries
+    flow_marker = ">"
+    fail_marker = "E"
+
+    def __init__(self, showlocals=False, style="long",
+                 abspath=True, tbfilter=True, funcargs=False):
+        self.showlocals = showlocals
+        self.style = style
+        self.tbfilter = tbfilter
+        self.funcargs = funcargs
+        self.abspath = abspath
+        self.astcache = {}
+
+    def _getindent(self, source):
+        # figure out indent for given source
+        try:
+            s = str(source.getstatement(len(source)-1))
+        except KeyboardInterrupt:
+            raise
+        except:
+            try:
+                s = str(source[-1])
+            except KeyboardInterrupt:
+                raise
+            except:
+                return 0
+        return 4 + (len(s) - len(s.lstrip()))
+
+    def _getentrysource(self, entry):
+        source = entry.getsource(self.astcache)
+        if source is not None:
+            source = source.deindent()
+        return source
+
+    def _saferepr(self, obj):
+        return py.io.saferepr(obj)
+
+    def repr_args(self, entry):
+        if self.funcargs:
+            args = []
+            for argname, argvalue in entry.frame.getargs(var=True):
+                args.append((argname, self._saferepr(argvalue)))
+            return ReprFuncArgs(args)
+
+    def get_source(self, source, line_index=-1, excinfo=None, short=False):
+        """ return formatted and marked up source lines. """
+        lines = []
+        if source is None or line_index >= len(source.lines):
+            source = py.code.Source("???")
+            line_index = 0
+        if line_index < 0:
+            line_index += len(source)
+        space_prefix = "    "
+        if short:
+            lines.append(space_prefix + source.lines[line_index].strip())
+        else:
+            for line in source.lines[:line_index]:
+                lines.append(space_prefix + line)
+            lines.append(self.flow_marker + "   " + source.lines[line_index])
+            for line in source.lines[line_index+1:]:
+                lines.append(space_prefix + line)
+        if excinfo is not None:
+            indent = 4 if short else self._getindent(source)
+            lines.extend(self.get_exconly(excinfo, indent=indent, markall=True))
+        return lines
+
+    def get_exconly(self, excinfo, indent=4, markall=False):
+        lines = []
+        indent = " " * indent
+        # get the real exception information out
+        exlines = excinfo.exconly(tryshort=True).split('\n')
+        failindent = self.fail_marker + indent[1:]
+        for line in exlines:
+            lines.append(failindent + line)
+            if not markall:
+                failindent = indent
+        return lines
+
+    def repr_locals(self, locals):
+        if self.showlocals:
+            lines = []
+            keys = [loc for loc in locals if loc[0] != "@"]
+            keys.sort()
+            for name in keys:
+                value = locals[name]
+                if name == '__builtins__':
+                    lines.append("__builtins__ = <builtins>")
+                else:
+                    # This formatting could all be handled by the
+                    # _repr() function, which is only reprlib.Repr in
+                    # disguise, so is very configurable.
+                    str_repr = self._saferepr(value)
+                    #if len(str_repr) < 70 or not isinstance(value,
+                    #                            (list, tuple, dict)):
+                    lines.append("%-10s = %s" %(name, str_repr))
+                    #else:
+                    #    self._line("%-10s =\\" % (name,))
+                    #    # XXX
+                    #    pprint.pprint(value, stream=self.excinfowriter)
+            return ReprLocals(lines)
+
+    def repr_traceback_entry(self, entry, excinfo=None):
+        source = self._getentrysource(entry)
+        if source is None:
+            source = py.code.Source("???")
+            line_index = 0
+        else:
+            # entry.getfirstlinesource() can be -1, should be 0 on jython
+            line_index = entry.lineno - max(entry.getfirstlinesource(), 0)
+
+        lines = []
+        style = entry._repr_style
+        if style is None:
+            style = self.style
+        if style in ("short", "long"):
+            short = style == "short"
+            reprargs = self.repr_args(entry) if not short else None
+            s = self.get_source(source, line_index, excinfo, short=short)
+            lines.extend(s)
+            if short:
+                message = "in %s" %(entry.name)
+            else:
+                message = excinfo and excinfo.typename or ""
+            path = self._makepath(entry.path)
+            filelocrepr = ReprFileLocation(path, entry.lineno+1, message)
+            localsrepr = None
+            if not short:
+                localsrepr =  self.repr_locals(entry.locals)
+            return ReprEntry(lines, reprargs, localsrepr, filelocrepr, style)
+        if excinfo:
+            lines.extend(self.get_exconly(excinfo, indent=4))
+        return ReprEntry(lines, None, None, None, style)
+
+    def _makepath(self, path):
+        if not self.abspath:
+            try:
+                np = py.path.local().bestrelpath(path)
+            except OSError:
+                return path
+            if len(np) < len(str(path)):
+                path = np
+        return path
+
+    def repr_traceback(self, excinfo):
+        traceback = excinfo.traceback
+        if self.tbfilter:
+            traceback = traceback.filter()
+        recursionindex = None
+        if excinfo.errisinstance(RuntimeError):
+            if "maximum recursion depth exceeded" in str(excinfo.value):
+                recursionindex = traceback.recursionindex()
+        last = traceback[-1]
+        entries = []
+        extraline = None
+        for index, entry in enumerate(traceback):
+            einfo = (last == entry) and excinfo or None
+            reprentry = self.repr_traceback_entry(entry, einfo)
+            entries.append(reprentry)
+            if index == recursionindex:
+                extraline = "!!! Recursion detected (same locals & position)"
+                break
+        return ReprTraceback(entries, extraline, style=self.style)
+
+    def repr_excinfo(self, excinfo):
+        reprtraceback = self.repr_traceback(excinfo)
+        reprcrash = excinfo._getreprcrash()
+        return ReprExceptionInfo(reprtraceback, reprcrash)
+
+class TerminalRepr:
+    def __str__(self):
+        s = self.__unicode__()
+        if sys.version_info[0] < 3:
+            s = s.encode('utf-8')
+        return s
+
+    def __unicode__(self):
+        # FYI this is called from pytest-xdist's serialization of exception
+        # information.
+        io = py.io.TextIO()
+        tw = py.io.TerminalWriter(file=io)
+        self.toterminal(tw)
+        return io.getvalue().strip()
+
+    def __repr__(self):
+        return "<%s instance at %0x>" %(self.__class__, id(self))
+
+
+class ReprExceptionInfo(TerminalRepr):
+    def __init__(self, reprtraceback, reprcrash):
+        self.reprtraceback = reprtraceback
+        self.reprcrash = reprcrash
+        self.sections = []
+
+    def addsection(self, name, content, sep="-"):
+        self.sections.append((name, content, sep))
+
+    def toterminal(self, tw):
+        self.reprtraceback.toterminal(tw)
+        for name, content, sep in self.sections:
+            tw.sep(sep, name)
+            tw.line(content)
+
+class ReprTraceback(TerminalRepr):
+    entrysep = "_ "
+
+    def __init__(self, reprentries, extraline, style):
+        self.reprentries = reprentries
+        self.extraline = extraline
+        self.style = style
+
+    def toterminal(self, tw):
+        # the entries might have different styles
+        last_style = None
+        for i, entry in enumerate(self.reprentries):
+            if entry.style == "long":
+                tw.line("")
+            entry.toterminal(tw)
+            if i < len(self.reprentries) - 1:
+                next_entry = self.reprentries[i+1]
+                if entry.style == "long" or \
+                   entry.style == "short" and next_entry.style == "long":
+                    tw.sep(self.entrysep)
+
+        if self.extraline:
+            tw.line(self.extraline)
+
+class ReprTracebackNative(ReprTraceback):
+    def __init__(self, tblines):
+        self.style = "native"
+        self.reprentries = [ReprEntryNative(tblines)]
+        self.extraline = None
+
+class ReprEntryNative(TerminalRepr):
+    style = "native"
+
+    def __init__(self, tblines):
+        self.lines = tblines
+
+    def toterminal(self, tw):
+        tw.write("".join(self.lines))
+
+class ReprEntry(TerminalRepr):
+    localssep = "_ "
+
+    def __init__(self, lines, reprfuncargs, reprlocals, filelocrepr, style):
+        self.lines = lines
+        self.reprfuncargs = reprfuncargs
+        self.reprlocals = reprlocals
+        self.reprfileloc = filelocrepr
+        self.style = style
+
+    def toterminal(self, tw):
+        if self.style == "short":
+            self.reprfileloc.toterminal(tw)
+            for line in self.lines:
+                red = line.startswith("E   ")
+                tw.line(line, bold=True, red=red)
+            #tw.line("")
+            return
+        if self.reprfuncargs:
+            self.reprfuncargs.toterminal(tw)
+        for line in self.lines:
+            red = line.startswith("E   ")
+            tw.line(line, bold=True, red=red)
+        if self.reprlocals:
+            #tw.sep(self.localssep, "Locals")
+            tw.line("")
+            self.reprlocals.toterminal(tw)
+        if self.reprfileloc:
+            if self.lines:
+                tw.line("")
+            self.reprfileloc.toterminal(tw)
+
+    def __str__(self):
+        return "%s\n%s\n%s" % ("\n".join(self.lines),
+                               self.reprlocals,
+                               self.reprfileloc)
+
+class ReprFileLocation(TerminalRepr):
+    def __init__(self, path, lineno, message):
+        self.path = str(path)
+        self.lineno = lineno
+        self.message = message
+
+    def toterminal(self, tw):
+        # filename and lineno output for each entry,
+        # using an output format that most editors unterstand
+        msg = self.message
+        i = msg.find("\n")
+        if i != -1:
+            msg = msg[:i]
+        tw.line("%s:%s: %s" %(self.path, self.lineno, msg))
+
+class ReprLocals(TerminalRepr):
+    def __init__(self, lines):
+        self.lines = lines
+
+    def toterminal(self, tw):
+        for line in self.lines:
+            tw.line(line)
+
+class ReprFuncArgs(TerminalRepr):
+    def __init__(self, args):
+        self.args = args
+
+    def toterminal(self, tw):
+        if self.args:
+            linesofar = ""
+            for name, value in self.args:
+                ns = "%s = %s" %(name, value)
+                if len(ns) + len(linesofar) + 2 > tw.fullwidth:
+                    if linesofar:
+                        tw.line(linesofar)
+                    linesofar =  ns
+                else:
+                    if linesofar:
+                        linesofar += ", " + ns
+                    else:
+                        linesofar = ns
+            if linesofar:
+                tw.line(linesofar)
+            tw.line("")
+
+
+
+oldbuiltins = {}
+
+def patch_builtins(assertion=True, compile=True):
+    """ put compile and AssertionError builtins to Python's builtins. """
+    if assertion:
+        from py._code import assertion
+        l = oldbuiltins.setdefault('AssertionError', [])
+        l.append(py.builtin.builtins.AssertionError)
+        py.builtin.builtins.AssertionError = assertion.AssertionError
+    if compile:
+        l = oldbuiltins.setdefault('compile', [])
+        l.append(py.builtin.builtins.compile)
+        py.builtin.builtins.compile = py.code.compile
+
+def unpatch_builtins(assertion=True, compile=True):
+    """ remove compile and AssertionError builtins from Python builtins. """
+    if assertion:
+        py.builtin.builtins.AssertionError = oldbuiltins['AssertionError'].pop()
+    if compile:
+        py.builtin.builtins.compile = oldbuiltins['compile'].pop()
+
+def getrawcode(obj, trycall=True):
+    """ return code object for given function. """
+    try:
+        return obj.__code__
+    except AttributeError:
+        obj = getattr(obj, 'im_func', obj)
+        obj = getattr(obj, 'func_code', obj)
+        obj = getattr(obj, 'f_code', obj)
+        obj = getattr(obj, '__code__', obj)
+        if trycall and not hasattr(obj, 'co_firstlineno'):
+            if hasattr(obj, '__call__') and not isclass(obj):
+                x = getrawcode(obj.__call__, trycall=False)
+                if hasattr(x, 'co_firstlineno'):
+                    return x
+        return obj
+
diff --git a/vllm/lib/python3.10/site-packages/py/_code/source.py b/vllm/lib/python3.10/site-packages/py/_code/source.py
new file mode 100644
index 0000000000000000000000000000000000000000..7fc7b23a96c32e603f1e678d5dad272e84a0e27d
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_code/source.py
@@ -0,0 +1,410 @@
+from __future__ import generators
+
+from bisect import bisect_right
+import sys
+import inspect, tokenize
+import py
+from types import ModuleType
+cpy_compile = compile
+
+try:
+    import _ast
+    from _ast import PyCF_ONLY_AST as _AST_FLAG
+except ImportError:
+    _AST_FLAG = 0
+    _ast = None
+
+
+class Source(object):
+    """ a immutable object holding a source code fragment,
+        possibly deindenting it.
+    """
+    _compilecounter = 0
+    def __init__(self, *parts, **kwargs):
+        self.lines = lines = []
+        de = kwargs.get('deindent', True)
+        rstrip = kwargs.get('rstrip', True)
+        for part in parts:
+            if not part:
+                partlines = []
+            if isinstance(part, Source):
+                partlines = part.lines
+            elif isinstance(part, (tuple, list)):
+                partlines = [x.rstrip("\n") for x in part]
+            elif isinstance(part, py.builtin._basestring):
+                partlines = part.split('\n')
+                if rstrip:
+                    while partlines:
+                        if partlines[-1].strip():
+                            break
+                        partlines.pop()
+            else:
+                partlines = getsource(part, deindent=de).lines
+            if de:
+                partlines = deindent(partlines)
+            lines.extend(partlines)
+
+    def __eq__(self, other):
+        try:
+            return self.lines == other.lines
+        except AttributeError:
+            if isinstance(other, str):
+                return str(self) == other
+            return False
+
+    def __getitem__(self, key):
+        if isinstance(key, int):
+            return self.lines[key]
+        else:
+            if key.step not in (None, 1):
+                raise IndexError("cannot slice a Source with a step")
+            return self.__getslice__(key.start, key.stop)
+
+    def __len__(self):
+        return len(self.lines)
+
+    def __getslice__(self, start, end):
+        newsource = Source()
+        newsource.lines = self.lines[start:end]
+        return newsource
+
+    def strip(self):
+        """ return new source object with trailing
+            and leading blank lines removed.
+        """
+        start, end = 0, len(self)
+        while start < end and not self.lines[start].strip():
+            start += 1
+        while end > start and not self.lines[end-1].strip():
+            end -= 1
+        source = Source()
+        source.lines[:] = self.lines[start:end]
+        return source
+
+    def putaround(self, before='', after='', indent=' ' * 4):
+        """ return a copy of the source object with
+            'before' and 'after' wrapped around it.
+        """
+        before = Source(before)
+        after = Source(after)
+        newsource = Source()
+        lines = [ (indent + line) for line in self.lines]
+        newsource.lines = before.lines + lines +  after.lines
+        return newsource
+
+    def indent(self, indent=' ' * 4):
+        """ return a copy of the source object with
+            all lines indented by the given indent-string.
+        """
+        newsource = Source()
+        newsource.lines = [(indent+line) for line in self.lines]
+        return newsource
+
+    def getstatement(self, lineno, assertion=False):
+        """ return Source statement which contains the
+            given linenumber (counted from 0).
+        """
+        start, end = self.getstatementrange(lineno, assertion)
+        return self[start:end]
+
+    def getstatementrange(self, lineno, assertion=False):
+        """ return (start, end) tuple which spans the minimal
+            statement region which containing the given lineno.
+        """
+        if not (0 <= lineno < len(self)):
+            raise IndexError("lineno out of range")
+        ast, start, end = getstatementrange_ast(lineno, self)
+        return start, end
+
+    def deindent(self, offset=None):
+        """ return a new source object deindented by offset.
+            If offset is None then guess an indentation offset from
+            the first non-blank line.  Subsequent lines which have a
+            lower indentation offset will be copied verbatim as
+            they are assumed to be part of multilines.
+        """
+        # XXX maybe use the tokenizer to properly handle multiline
+        #     strings etc.pp?
+        newsource = Source()
+        newsource.lines[:] = deindent(self.lines, offset)
+        return newsource
+
+    def isparseable(self, deindent=True):
+        """ return True if source is parseable, heuristically
+            deindenting it by default.
+        """
+        try:
+            import parser
+        except ImportError:
+            syntax_checker = lambda x: compile(x, 'asd', 'exec')
+        else:
+            syntax_checker = parser.suite
+
+        if deindent:
+            source = str(self.deindent())
+        else:
+            source = str(self)
+        try:
+            #compile(source+'\n', "x", "exec")
+            syntax_checker(source+'\n')
+        except KeyboardInterrupt:
+            raise
+        except Exception:
+            return False
+        else:
+            return True
+
+    def __str__(self):
+        return "\n".join(self.lines)
+
+    def compile(self, filename=None, mode='exec',
+                flag=generators.compiler_flag,
+                dont_inherit=0, _genframe=None):
+        """ return compiled code object. if filename is None
+            invent an artificial filename which displays
+            the source/line position of the caller frame.
+        """
+        if not filename or py.path.local(filename).check(file=0):
+            if _genframe is None:
+                _genframe = sys._getframe(1) # the caller
+            fn,lineno = _genframe.f_code.co_filename, _genframe.f_lineno
+            base = "<%d-codegen " % self._compilecounter
+            self.__class__._compilecounter += 1
+            if not filename:
+                filename = base + '%s:%d>' % (fn, lineno)
+            else:
+                filename = base + '%r %s:%d>' % (filename, fn, lineno)
+        source = "\n".join(self.lines) + '\n'
+        try:
+            co = cpy_compile(source, filename, mode, flag)
+        except SyntaxError:
+            ex = sys.exc_info()[1]
+            # re-represent syntax errors from parsing python strings
+            msglines = self.lines[:ex.lineno]
+            if ex.offset:
+                msglines.append(" "*ex.offset + '^')
+            msglines.append("(code was compiled probably from here: %s)" % filename)
+            newex = SyntaxError('\n'.join(msglines))
+            newex.offset = ex.offset
+            newex.lineno = ex.lineno
+            newex.text = ex.text
+            raise newex
+        else:
+            if flag & _AST_FLAG:
+                return co
+            lines = [(x + "\n") for x in self.lines]
+            import linecache
+            linecache.cache[filename] = (1, None, lines, filename)
+            return co
+
+#
+# public API shortcut functions
+#
+
+def compile_(source, filename=None, mode='exec', flags=
+            generators.compiler_flag, dont_inherit=0):
+    """ compile the given source to a raw code object,
+        and maintain an internal cache which allows later
+        retrieval of the source code for the code object
+        and any recursively created code objects.
+    """
+    if _ast is not None and isinstance(source, _ast.AST):
+        # XXX should Source support having AST?
+        return cpy_compile(source, filename, mode, flags, dont_inherit)
+    _genframe = sys._getframe(1) # the caller
+    s = Source(source)
+    co = s.compile(filename, mode, flags, _genframe=_genframe)
+    return co
+
+
+def getfslineno(obj):
+    """ Return source location (path, lineno) for the given object.
+    If the source cannot be determined return ("", -1)
+    """
+    try:
+        code = py.code.Code(obj)
+    except TypeError:
+        try:
+            fn = (inspect.getsourcefile(obj) or
+                  inspect.getfile(obj))
+        except TypeError:
+            return "", -1
+
+        fspath = fn and py.path.local(fn) or None
+        lineno = -1
+        if fspath:
+            try:
+                _, lineno = findsource(obj)
+            except IOError:
+                pass
+    else:
+        fspath = code.path
+        lineno = code.firstlineno
+    assert isinstance(lineno, int)
+    return fspath, lineno
+
+#
+# helper functions
+#
+
+def findsource(obj):
+    try:
+        sourcelines, lineno = inspect.findsource(obj)
+    except py.builtin._sysex:
+        raise
+    except:
+        return None, -1
+    source = Source()
+    source.lines = [line.rstrip() for line in sourcelines]
+    return source, lineno
+
+def getsource(obj, **kwargs):
+    obj = py.code.getrawcode(obj)
+    try:
+        strsrc = inspect.getsource(obj)
+    except IndentationError:
+        strsrc = "\"Buggy python version consider upgrading, cannot get source\""
+    assert isinstance(strsrc, str)
+    return Source(strsrc, **kwargs)
+
+def deindent(lines, offset=None):
+    if offset is None:
+        for line in lines:
+            line = line.expandtabs()
+            s = line.lstrip()
+            if s:
+                offset = len(line)-len(s)
+                break
+        else:
+            offset = 0
+    if offset == 0:
+        return list(lines)
+    newlines = []
+    def readline_generator(lines):
+        for line in lines:
+            yield line + '\n'
+        while True:
+            yield ''
+
+    it = readline_generator(lines)
+
+    try:
+        for _, _, (sline, _), (eline, _), _ in tokenize.generate_tokens(lambda: next(it)):
+            if sline > len(lines):
+                break # End of input reached
+            if sline > len(newlines):
+                line = lines[sline - 1].expandtabs()
+                if line.lstrip() and line[:offset].isspace():
+                    line = line[offset:] # Deindent
+                newlines.append(line)
+
+            for i in range(sline, eline):
+                # Don't deindent continuing lines of
+                # multiline tokens (i.e. multiline strings)
+                newlines.append(lines[i])
+    except (IndentationError, tokenize.TokenError):
+        pass
+    # Add any lines we didn't see. E.g. if an exception was raised.
+    newlines.extend(lines[len(newlines):])
+    return newlines
+
+
+def get_statement_startend2(lineno, node):
+    import ast
+    # flatten all statements and except handlers into one lineno-list
+    # AST's line numbers start indexing at 1
+    l = []
+    for x in ast.walk(node):
+        if isinstance(x, _ast.stmt) or isinstance(x, _ast.ExceptHandler):
+            l.append(x.lineno - 1)
+            for name in "finalbody", "orelse":
+                val = getattr(x, name, None)
+                if val:
+                    # treat the finally/orelse part as its own statement
+                    l.append(val[0].lineno - 1 - 1)
+    l.sort()
+    insert_index = bisect_right(l, lineno)
+    start = l[insert_index - 1]
+    if insert_index >= len(l):
+        end = None
+    else:
+        end = l[insert_index]
+    return start, end
+
+
+def getstatementrange_ast(lineno, source, assertion=False, astnode=None):
+    if astnode is None:
+        content = str(source)
+        try:
+            astnode = compile(content, "source", "exec", 1024)  # 1024 for AST
+        except ValueError:
+            start, end = getstatementrange_old(lineno, source, assertion)
+            return None, start, end
+    start, end = get_statement_startend2(lineno, astnode)
+    # we need to correct the end:
+    # - ast-parsing strips comments
+    # - there might be empty lines
+    # - we might have lesser indented code blocks at the end
+    if end is None:
+        end = len(source.lines)
+
+    if end > start + 1:
+        # make sure we don't span differently indented code blocks
+        # by using the BlockFinder helper used which inspect.getsource() uses itself
+        block_finder = inspect.BlockFinder()
+        # if we start with an indented line, put blockfinder to "started" mode
+        block_finder.started = source.lines[start][0].isspace()
+        it = ((x + "\n") for x in source.lines[start:end])
+        try:
+            for tok in tokenize.generate_tokens(lambda: next(it)):
+                block_finder.tokeneater(*tok)
+        except (inspect.EndOfBlock, IndentationError):
+            end = block_finder.last + start
+        except Exception:
+            pass
+
+    # the end might still point to a comment or empty line, correct it
+    while end:
+        line = source.lines[end - 1].lstrip()
+        if line.startswith("#") or not line:
+            end -= 1
+        else:
+            break
+    return astnode, start, end
+
+
+def getstatementrange_old(lineno, source, assertion=False):
+    """ return (start, end) tuple which spans the minimal
+        statement region which containing the given lineno.
+        raise an IndexError if no such statementrange can be found.
+    """
+    # XXX this logic is only used on python2.4 and below
+    # 1. find the start of the statement
+    from codeop import compile_command
+    for start in range(lineno, -1, -1):
+        if assertion:
+            line = source.lines[start]
+            # the following lines are not fully tested, change with care
+            if 'super' in line and 'self' in line and '__init__' in line:
+                raise IndexError("likely a subclass")
+            if "assert" not in line and "raise" not in line:
+                continue
+        trylines = source.lines[start:lineno+1]
+        # quick hack to prepare parsing an indented line with
+        # compile_command() (which errors on "return" outside defs)
+        trylines.insert(0, 'def xxx():')
+        trysource = '\n '.join(trylines)
+        #              ^ space here
+        try:
+            compile_command(trysource)
+        except (SyntaxError, OverflowError, ValueError):
+            continue
+
+        # 2. find the end of the statement
+        for end in range(lineno+1, len(source)+1):
+            trysource = source[start:end]
+            if trysource.isparseable():
+                return start, end
+    raise SyntaxError("no valid source range around line %d " % (lineno,))
+
+
diff --git a/vllm/lib/python3.10/site-packages/py/_io/__init__.py b/vllm/lib/python3.10/site-packages/py/_io/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..835f01f3ab9dcb656dce1e580f0d98d7b8abfe3a
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_io/__init__.py
@@ -0,0 +1 @@
+""" input/output helping """
diff --git a/vllm/lib/python3.10/site-packages/py/_io/__pycache__/__init__.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c8490f7c53a6e9c84e5867dab412bf3e9ae289a4
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/__init__.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_io/__pycache__/capture.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/capture.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..20d97f5d95c2cfb8685b47d6462337ba692f74a2
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/capture.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_io/__pycache__/saferepr.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/saferepr.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..19b397284fe1570ce020ff99aa84d23de1f8857e
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/saferepr.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_io/__pycache__/terminalwriter.cpython-310.pyc b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/terminalwriter.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..14c609a96ffb85f2bf194dbbb6852d2fe94b232a
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/py/_io/__pycache__/terminalwriter.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/py/_io/capture.py b/vllm/lib/python3.10/site-packages/py/_io/capture.py
new file mode 100644
index 0000000000000000000000000000000000000000..cacf2fa71a104bd3a4c25582bded149faa0c1a9f
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_io/capture.py
@@ -0,0 +1,371 @@
+import os
+import sys
+import py
+import tempfile
+
+try:
+    from io import StringIO
+except ImportError:
+    from StringIO import StringIO
+
+if sys.version_info < (3,0):
+    class TextIO(StringIO):
+        def write(self, data):
+            if not isinstance(data, unicode):
+                data = unicode(data, getattr(self, '_encoding', 'UTF-8'), 'replace')
+            return StringIO.write(self, data)
+else:
+    TextIO = StringIO
+
+try:
+    from io import BytesIO
+except ImportError:
+    class BytesIO(StringIO):
+        def write(self, data):
+            if isinstance(data, unicode):
+                raise TypeError("not a byte value: %r" %(data,))
+            return StringIO.write(self, data)
+
+patchsysdict = {0: 'stdin', 1: 'stdout', 2: 'stderr'}
+
+class FDCapture:
+    """ Capture IO to/from a given os-level filedescriptor. """
+
+    def __init__(self, targetfd, tmpfile=None, now=True, patchsys=False):
+        """ save targetfd descriptor, and open a new
+            temporary file there.  If no tmpfile is
+            specified a tempfile.Tempfile() will be opened
+            in text mode.
+        """
+        self.targetfd = targetfd
+        if tmpfile is None and targetfd != 0:
+            f = tempfile.TemporaryFile('wb+')
+            tmpfile = dupfile(f, encoding="UTF-8")
+            f.close()
+        self.tmpfile = tmpfile
+        self._savefd = os.dup(self.targetfd)
+        if patchsys:
+            self._oldsys = getattr(sys, patchsysdict[targetfd])
+        if now:
+            self.start()
+
+    def start(self):
+        try:
+            os.fstat(self._savefd)
+        except OSError:
+            raise ValueError("saved filedescriptor not valid, "
+                "did you call start() twice?")
+        if self.targetfd == 0 and not self.tmpfile:
+            fd = os.open(devnullpath, os.O_RDONLY)
+            os.dup2(fd, 0)
+            os.close(fd)
+            if hasattr(self, '_oldsys'):
+                setattr(sys, patchsysdict[self.targetfd], DontReadFromInput())
+        else:
+            os.dup2(self.tmpfile.fileno(), self.targetfd)
+            if hasattr(self, '_oldsys'):
+                setattr(sys, patchsysdict[self.targetfd], self.tmpfile)
+
+    def done(self):
+        """ unpatch and clean up, returns the self.tmpfile (file object)
+        """
+        os.dup2(self._savefd, self.targetfd)
+        os.close(self._savefd)
+        if self.targetfd != 0:
+            self.tmpfile.seek(0)
+        if hasattr(self, '_oldsys'):
+            setattr(sys, patchsysdict[self.targetfd], self._oldsys)
+        return self.tmpfile
+
+    def writeorg(self, data):
+        """ write a string to the original file descriptor
+        """
+        tempfp = tempfile.TemporaryFile()
+        try:
+            os.dup2(self._savefd, tempfp.fileno())
+            tempfp.write(data)
+        finally:
+            tempfp.close()
+
+
+def dupfile(f, mode=None, buffering=0, raising=False, encoding=None):
+    """ return a new open file object that's a duplicate of f
+
+        mode is duplicated if not given, 'buffering' controls
+        buffer size (defaulting to no buffering) and 'raising'
+        defines whether an exception is raised when an incompatible
+        file object is passed in (if raising is False, the file
+        object itself will be returned)
+    """
+    try:
+        fd = f.fileno()
+        mode = mode or f.mode
+    except AttributeError:
+        if raising:
+            raise
+        return f
+    newfd = os.dup(fd)
+    if sys.version_info >= (3,0):
+        if encoding is not None:
+            mode = mode.replace("b", "")
+            buffering = True
+        return os.fdopen(newfd, mode, buffering, encoding, closefd=True)
+    else:
+        f = os.fdopen(newfd, mode, buffering)
+        if encoding is not None:
+            return EncodedFile(f, encoding)
+        return f
+
+class EncodedFile(object):
+    def __init__(self, _stream, encoding):
+        self._stream = _stream
+        self.encoding = encoding
+
+    def write(self, obj):
+        if isinstance(obj, unicode):
+            obj = obj.encode(self.encoding)
+        elif isinstance(obj, str):
+            pass
+        else:
+            obj = str(obj)
+        self._stream.write(obj)
+
+    def writelines(self, linelist):
+        data = ''.join(linelist)
+        self.write(data)
+
+    def __getattr__(self, name):
+        return getattr(self._stream, name)
+
+class Capture(object):
+    def call(cls, func, *args, **kwargs):
+        """ return a (res, out, err) tuple where
+            out and err represent the output/error output
+            during function execution.
+            call the given function with args/kwargs
+            and capture output/error during its execution.
+        """
+        so = cls()
+        try:
+            res = func(*args, **kwargs)
+        finally:
+            out, err = so.reset()
+        return res, out, err
+    call = classmethod(call)
+
+    def reset(self):
+        """ reset sys.stdout/stderr and return captured output as strings. """
+        if hasattr(self, '_reset'):
+            raise ValueError("was already reset")
+        self._reset = True
+        outfile, errfile = self.done(save=False)
+        out, err = "", ""
+        if outfile and not outfile.closed:
+            out = outfile.read()
+            outfile.close()
+        if errfile and errfile != outfile and not errfile.closed:
+            err = errfile.read()
+            errfile.close()
+        return out, err
+
+    def suspend(self):
+        """ return current snapshot captures, memorize tempfiles. """
+        outerr = self.readouterr()
+        outfile, errfile = self.done()
+        return outerr
+
+
+class StdCaptureFD(Capture):
+    """ This class allows to capture writes to FD1 and FD2
+        and may connect a NULL file to FD0 (and prevent
+        reads from sys.stdin).  If any of the 0,1,2 file descriptors
+        is invalid it will not be captured.
+    """
+    def __init__(self, out=True, err=True, mixed=False,
+        in_=True, patchsys=True, now=True):
+        self._options = {
+            "out": out,
+            "err": err,
+            "mixed": mixed,
+            "in_": in_,
+            "patchsys": patchsys,
+            "now": now,
+        }
+        self._save()
+        if now:
+            self.startall()
+
+    def _save(self):
+        in_ = self._options['in_']
+        out = self._options['out']
+        err = self._options['err']
+        mixed = self._options['mixed']
+        patchsys = self._options['patchsys']
+        if in_:
+            try:
+                self.in_ = FDCapture(0, tmpfile=None, now=False,
+                    patchsys=patchsys)
+            except OSError:
+                pass
+        if out:
+            tmpfile = None
+            if hasattr(out, 'write'):
+                tmpfile = out
+            try:
+                self.out = FDCapture(1, tmpfile=tmpfile,
+                           now=False, patchsys=patchsys)
+                self._options['out'] = self.out.tmpfile
+            except OSError:
+                pass
+        if err:
+            if out and mixed:
+                tmpfile = self.out.tmpfile
+            elif hasattr(err, 'write'):
+                tmpfile = err
+            else:
+                tmpfile = None
+            try:
+                self.err = FDCapture(2, tmpfile=tmpfile,
+                           now=False, patchsys=patchsys)
+                self._options['err'] = self.err.tmpfile
+            except OSError:
+                pass
+
+    def startall(self):
+        if hasattr(self, 'in_'):
+            self.in_.start()
+        if hasattr(self, 'out'):
+            self.out.start()
+        if hasattr(self, 'err'):
+            self.err.start()
+
+    def resume(self):
+        """ resume capturing with original temp files. """
+        self.startall()
+
+    def done(self, save=True):
+        """ return (outfile, errfile) and stop capturing. """
+        outfile = errfile = None
+        if hasattr(self, 'out') and not self.out.tmpfile.closed:
+            outfile = self.out.done()
+        if hasattr(self, 'err') and not self.err.tmpfile.closed:
+            errfile = self.err.done()
+        if hasattr(self, 'in_'):
+            tmpfile = self.in_.done()
+        if save:
+            self._save()
+        return outfile, errfile
+
+    def readouterr(self):
+        """ return snapshot value of stdout/stderr capturings. """
+        if hasattr(self, "out"):
+            out = self._readsnapshot(self.out.tmpfile)
+        else:
+            out = ""
+        if hasattr(self, "err"):
+            err = self._readsnapshot(self.err.tmpfile)
+        else:
+            err = ""
+        return out, err
+
+    def _readsnapshot(self, f):
+        f.seek(0)
+        res = f.read()
+        enc = getattr(f, "encoding", None)
+        if enc:
+            res = py.builtin._totext(res, enc, "replace")
+        f.truncate(0)
+        f.seek(0)
+        return res
+
+
+class StdCapture(Capture):
+    """ This class allows to capture writes to sys.stdout|stderr "in-memory"
+        and will raise errors on tries to read from sys.stdin. It only
+        modifies sys.stdout|stderr|stdin attributes and does not
+        touch underlying File Descriptors (use StdCaptureFD for that).
+    """
+    def __init__(self, out=True, err=True, in_=True, mixed=False, now=True):
+        self._oldout = sys.stdout
+        self._olderr = sys.stderr
+        self._oldin  = sys.stdin
+        if out and not hasattr(out, 'file'):
+            out = TextIO()
+        self.out = out
+        if err:
+            if mixed:
+                err = out
+            elif not hasattr(err, 'write'):
+                err = TextIO()
+        self.err = err
+        self.in_ = in_
+        if now:
+            self.startall()
+
+    def startall(self):
+        if self.out:
+            sys.stdout = self.out
+        if self.err:
+            sys.stderr = self.err
+        if self.in_:
+            sys.stdin  = self.in_  = DontReadFromInput()
+
+    def done(self, save=True):
+        """ return (outfile, errfile) and stop capturing. """
+        outfile = errfile = None
+        if self.out and not self.out.closed:
+            sys.stdout = self._oldout
+            outfile = self.out
+            outfile.seek(0)
+        if self.err and not self.err.closed:
+            sys.stderr = self._olderr
+            errfile = self.err
+            errfile.seek(0)
+        if self.in_:
+            sys.stdin = self._oldin
+        return outfile, errfile
+
+    def resume(self):
+        """ resume capturing with original temp files. """
+        self.startall()
+
+    def readouterr(self):
+        """ return snapshot value of stdout/stderr capturings. """
+        out = err = ""
+        if self.out:
+            out = self.out.getvalue()
+            self.out.truncate(0)
+            self.out.seek(0)
+        if self.err:
+            err = self.err.getvalue()
+            self.err.truncate(0)
+            self.err.seek(0)
+        return out, err
+
+class DontReadFromInput:
+    """Temporary stub class.  Ideally when stdin is accessed, the
+    capturing should be turned off, with possibly all data captured
+    so far sent to the screen.  This should be configurable, though,
+    because in automated test runs it is better to crash than
+    hang indefinitely.
+    """
+    def read(self, *args):
+        raise IOError("reading from stdin while output is captured")
+    readline = read
+    readlines = read
+    __iter__ = read
+
+    def fileno(self):
+        raise ValueError("redirected Stdin is pseudofile, has no fileno()")
+    def isatty(self):
+        return False
+    def close(self):
+        pass
+
+try:
+    devnullpath = os.devnull
+except AttributeError:
+    if os.name == 'nt':
+        devnullpath = 'NUL'
+    else:
+        devnullpath = '/dev/null'
diff --git a/vllm/lib/python3.10/site-packages/py/_io/saferepr.py b/vllm/lib/python3.10/site-packages/py/_io/saferepr.py
new file mode 100644
index 0000000000000000000000000000000000000000..8518290efddecdc8524c642abbd5aba76dada44c
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_io/saferepr.py
@@ -0,0 +1,71 @@
+import py
+import sys
+
+builtin_repr = repr
+
+reprlib = py.builtin._tryimport('repr', 'reprlib')
+
+class SafeRepr(reprlib.Repr):
+    """ subclass of repr.Repr that limits the resulting size of repr()
+        and includes information on exceptions raised during the call.
+    """
+    def repr(self, x):
+        return self._callhelper(reprlib.Repr.repr, self, x)
+
+    def repr_unicode(self, x, level):
+        # Strictly speaking wrong on narrow builds
+        def repr(u):
+            if "'" not in u:
+                return py.builtin._totext("'%s'") % u
+            elif '"' not in u:
+                return py.builtin._totext('"%s"') % u
+            else:
+                return py.builtin._totext("'%s'") % u.replace("'", r"\'")
+        s = repr(x[:self.maxstring])
+        if len(s) > self.maxstring:
+            i = max(0, (self.maxstring-3)//2)
+            j = max(0, self.maxstring-3-i)
+            s = repr(x[:i] + x[len(x)-j:])
+            s = s[:i] + '...' + s[len(s)-j:]
+        return s
+
+    def repr_instance(self, x, level):
+        return self._callhelper(builtin_repr, x)
+
+    def _callhelper(self, call, x, *args):
+        try:
+            # Try the vanilla repr and make sure that the result is a string
+            s = call(x, *args)
+        except py.builtin._sysex:
+            raise
+        except:
+            cls, e, tb = sys.exc_info()
+            exc_name = getattr(cls, '__name__', 'unknown')
+            try:
+                exc_info = str(e)
+            except py.builtin._sysex:
+                raise
+            except:
+                exc_info = 'unknown'
+            return '<[%s("%s") raised in repr()] %s object at 0x%x>' % (
+                exc_name, exc_info, x.__class__.__name__, id(x))
+        else:
+            if len(s) > self.maxsize:
+                i = max(0, (self.maxsize-3)//2)
+                j = max(0, self.maxsize-3-i)
+                s = s[:i] + '...' + s[len(s)-j:]
+            return s
+
+def saferepr(obj, maxsize=240):
+    """ return a size-limited safe repr-string for the given object.
+    Failing __repr__ functions of user instances will be represented
+    with a short exception info and 'saferepr' generally takes
+    care to never raise exceptions itself.  This function is a wrapper
+    around the Repr/reprlib functionality of the standard 2.6 lib.
+    """
+    # review exception handling
+    srepr = SafeRepr()
+    srepr.maxstring = maxsize
+    srepr.maxsize = maxsize
+    srepr.maxother = 160
+    return srepr.repr(obj)
diff --git a/vllm/lib/python3.10/site-packages/py/_io/terminalwriter.py b/vllm/lib/python3.10/site-packages/py/_io/terminalwriter.py
new file mode 100644
index 0000000000000000000000000000000000000000..442ca2395e0a15e8cddf2da9565d02b2c807faaf
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_io/terminalwriter.py
@@ -0,0 +1,423 @@
+"""
+
+Helper functions for writing to terminals and files.
+
+"""
+
+
+import sys, os, unicodedata
+import py
+py3k = sys.version_info[0] >= 3
+py33 = sys.version_info >= (3, 3)
+from py.builtin import text, bytes
+
+win32_and_ctypes = False
+colorama = None
+if sys.platform == "win32":
+    try:
+        import colorama
+    except ImportError:
+        try:
+            import ctypes
+            win32_and_ctypes = True
+        except ImportError:
+            pass
+
+
+def _getdimensions():
+    if py33:
+        import shutil
+        size = shutil.get_terminal_size()
+        return size.lines, size.columns
+    else:
+        import termios, fcntl, struct
+        call = fcntl.ioctl(1, termios.TIOCGWINSZ, "\000" * 8)
+        height, width = struct.unpack("hhhh", call)[:2]
+        return height, width
+
+
+def get_terminal_width():
+    width = 0
+    try:
+        _, width = _getdimensions()
+    except py.builtin._sysex:
+        raise
+    except:
+        # pass to fallback below
+        pass
+
+    if width == 0:
+        # FALLBACK:
+        # * some exception happened
+        # * or this is emacs terminal which reports (0,0)
+        width = int(os.environ.get('COLUMNS', 80))
+
+    # XXX the windows getdimensions may be bogus, let's sanify a bit
+    if width < 40:
+        width = 80
+    return width
+
+terminal_width = get_terminal_width()
+
+char_width = {
+    'A': 1,   # "Ambiguous"
+    'F': 2,   # Fullwidth
+    'H': 1,   # Halfwidth
+    'N': 1,   # Neutral
+    'Na': 1,  # Narrow
+    'W': 2,   # Wide
+}
+
+
+def get_line_width(text):
+    text = unicodedata.normalize('NFC', text)
+    return sum(char_width.get(unicodedata.east_asian_width(c), 1) for c in text)
+
+
+# XXX unify with _escaped func below
+def ansi_print(text, esc, file=None, newline=True, flush=False):
+    if file is None:
+        file = sys.stderr
+    text = text.rstrip()
+    if esc and not isinstance(esc, tuple):
+        esc = (esc,)
+    if esc and sys.platform != "win32" and file.isatty():
+        text = (''.join(['\x1b[%sm' % cod for cod in esc])  +
+                text +
+                '\x1b[0m')     # ANSI color code "reset"
+    if newline:
+        text += '\n'
+
+    if esc and win32_and_ctypes and file.isatty():
+        if 1 in esc:
+            bold = True
+            esc = tuple([x for x in esc if x != 1])
+        else:
+            bold = False
+        esctable = {()   : FOREGROUND_WHITE,                 # normal
+                    (31,): FOREGROUND_RED,                   # red
+                    (32,): FOREGROUND_GREEN,                 # green
+                    (33,): FOREGROUND_GREEN|FOREGROUND_RED,  # yellow
+                    (34,): FOREGROUND_BLUE,                  # blue
+                    (35,): FOREGROUND_BLUE|FOREGROUND_RED,   # purple
+                    (36,): FOREGROUND_BLUE|FOREGROUND_GREEN, # cyan
+                    (37,): FOREGROUND_WHITE,                 # white
+                    (39,): FOREGROUND_WHITE,                 # reset
+                    }
+        attr = esctable.get(esc, FOREGROUND_WHITE)
+        if bold:
+            attr |= FOREGROUND_INTENSITY
+        STD_OUTPUT_HANDLE = -11
+        STD_ERROR_HANDLE = -12
+        if file is sys.stderr:
+            handle = GetStdHandle(STD_ERROR_HANDLE)
+        else:
+            handle = GetStdHandle(STD_OUTPUT_HANDLE)
+        oldcolors = GetConsoleInfo(handle).wAttributes
+        attr |= (oldcolors & 0x0f0)
+        SetConsoleTextAttribute(handle, attr)
+        while len(text) > 32768:
+            file.write(text[:32768])
+            text = text[32768:]
+        if text:
+            file.write(text)
+        SetConsoleTextAttribute(handle, oldcolors)
+    else:
+        file.write(text)
+
+    if flush:
+        file.flush()
+
+def should_do_markup(file):
+    if os.environ.get('PY_COLORS') == '1':
+        return True
+    if os.environ.get('PY_COLORS') == '0':
+        return False
+    if 'NO_COLOR' in os.environ:
+        return False
+    return hasattr(file, 'isatty') and file.isatty() \
+           and os.environ.get('TERM') != 'dumb' \
+           and not (sys.platform.startswith('java') and os._name == 'nt')
+
+class TerminalWriter(object):
+    _esctable = dict(black=30, red=31, green=32, yellow=33,
+                     blue=34, purple=35, cyan=36, white=37,
+                     Black=40, Red=41, Green=42, Yellow=43,
+                     Blue=44, Purple=45, Cyan=46, White=47,
+                     bold=1, light=2, blink=5, invert=7)
+
+    # XXX deprecate stringio argument
+    def __init__(self, file=None, stringio=False, encoding=None):
+        if file is None:
+            if stringio:
+                self.stringio = file = py.io.TextIO()
+            else:
+                from sys import stdout as file
+        elif py.builtin.callable(file) and not (
+             hasattr(file, "write") and hasattr(file, "flush")):
+            file = WriteFile(file, encoding=encoding)
+        if hasattr(file, "isatty") and file.isatty() and colorama:
+            file = colorama.AnsiToWin32(file).stream
+        self.encoding = encoding or getattr(file, 'encoding', "utf-8")
+        self._file = file
+        self.hasmarkup = should_do_markup(file)
+        self._lastlen = 0
+        self._chars_on_current_line = 0
+        self._width_of_current_line = 0
+
+    @property
+    def fullwidth(self):
+        if hasattr(self, '_terminal_width'):
+            return self._terminal_width
+        return get_terminal_width()
+
+    @fullwidth.setter
+    def fullwidth(self, value):
+        self._terminal_width = value
+
+    @property
+    def chars_on_current_line(self):
+        """Return the number of characters written so far in the current line.
+
+        Please note that this count does not produce correct results after a reline() call,
+        see #164.
+
+        .. versionadded:: 1.5.0
+
+        :rtype: int
+        """
+        return self._chars_on_current_line
+
+    @property
+    def width_of_current_line(self):
+        """Return an estimate of the width so far in the current line.
+
+        .. versionadded:: 1.6.0
+
+        :rtype: int
+        """
+        return self._width_of_current_line
+
+    def _escaped(self, text, esc):
+        if esc and self.hasmarkup:
+            text = (''.join(['\x1b[%sm' % cod for cod in esc])  +
+                text +'\x1b[0m')
+        return text
+
+    def markup(self, text, **kw):
+        esc = []
+        for name in kw:
+            if name not in self._esctable:
+                raise ValueError("unknown markup: %r" %(name,))
+            if kw[name]:
+                esc.append(self._esctable[name])
+        return self._escaped(text, tuple(esc))
+
+    def sep(self, sepchar, title=None, fullwidth=None, **kw):
+        if fullwidth is None:
+            fullwidth = self.fullwidth
+        # the goal is to have the line be as long as possible
+        # under the condition that len(line) <= fullwidth
+        if sys.platform == "win32":
+            # if we print in the last column on windows we are on a
+            # new line but there is no way to verify/neutralize this
+            # (we may not know the exact line width)
+            # so let's be defensive to avoid empty lines in the output
+            fullwidth -= 1
+        if title is not None:
+            # we want 2 + 2*len(fill) + len(title) <= fullwidth
+            # i.e.    2 + 2*len(sepchar)*N + len(title) <= fullwidth
+            #         2*len(sepchar)*N <= fullwidth - len(title) - 2
+            #         N <= (fullwidth - len(title) - 2) // (2*len(sepchar))
+            N = max((fullwidth - len(title) - 2) // (2*len(sepchar)), 1)
+            fill = sepchar * N
+            line = "%s %s %s" % (fill, title, fill)
+        else:
+            # we want len(sepchar)*N <= fullwidth
+            # i.e.    N <= fullwidth // len(sepchar)
+            line = sepchar * (fullwidth // len(sepchar))
+        # in some situations there is room for an extra sepchar at the right,
+        # in particular if we consider that with a sepchar like "_ " the
+        # trailing space is not important at the end of the line
+        if len(line) + len(sepchar.rstrip()) <= fullwidth:
+            line += sepchar.rstrip()
+
+        self.line(line, **kw)
+
+    def write(self, msg, **kw):
+        if msg:
+            if not isinstance(msg, (bytes, text)):
+                msg = text(msg)
+
+            self._update_chars_on_current_line(msg)
+
+            if self.hasmarkup and kw:
+                markupmsg = self.markup(msg, **kw)
+            else:
+                markupmsg = msg
+            write_out(self._file, markupmsg)
+
+    def _update_chars_on_current_line(self, text_or_bytes):
+        newline = b'\n' if isinstance(text_or_bytes, bytes) else '\n'
+        current_line = text_or_bytes.rsplit(newline, 1)[-1]
+        if isinstance(current_line, bytes):
+            current_line = current_line.decode('utf-8', errors='replace')
+        if newline in text_or_bytes:
+            self._chars_on_current_line = len(current_line)
+            self._width_of_current_line = get_line_width(current_line)
+        else:
+            self._chars_on_current_line += len(current_line)
+            self._width_of_current_line += get_line_width(current_line)
+
+    def line(self, s='', **kw):
+        self.write(s, **kw)
+        self._checkfill(s)
+        self.write('\n')
+
+    def reline(self, line, **kw):
+        if not self.hasmarkup:
+            raise ValueError("cannot use rewrite-line without terminal")
+        self.write(line, **kw)
+        self._checkfill(line)
+        self.write('\r')
+        self._lastlen = len(line)
+
+    def _checkfill(self, line):
+        diff2last = self._lastlen - len(line)
+        if diff2last > 0:
+            self.write(" " * diff2last)
+
+class Win32ConsoleWriter(TerminalWriter):
+    def write(self, msg, **kw):
+        if msg:
+            if not isinstance(msg, (bytes, text)):
+                msg = text(msg)
+
+            self._update_chars_on_current_line(msg)
+
+            oldcolors = None
+            if self.hasmarkup and kw:
+                handle = GetStdHandle(STD_OUTPUT_HANDLE)
+                oldcolors = GetConsoleInfo(handle).wAttributes
+                default_bg = oldcolors & 0x00F0
+                attr = default_bg
+                if kw.pop('bold', False):
+                    attr |= FOREGROUND_INTENSITY
+
+                if kw.pop('red', False):
+                    attr |= FOREGROUND_RED
+                elif kw.pop('blue', False):
+                    attr |= FOREGROUND_BLUE
+                elif kw.pop('green', False):
+                    attr |= FOREGROUND_GREEN
+                elif kw.pop('yellow', False):
+                    attr |= FOREGROUND_GREEN|FOREGROUND_RED
+                else:
+                    attr |= oldcolors & 0x0007
+
+                SetConsoleTextAttribute(handle, attr)
+            write_out(self._file, msg)
+            if oldcolors:
+                SetConsoleTextAttribute(handle, oldcolors)
+
+class WriteFile(object):
+    def __init__(self, writemethod, encoding=None):
+        self.encoding = encoding
+        self._writemethod = writemethod
+
+    def write(self, data):
+        if self.encoding:
+            data = data.encode(self.encoding, "replace")
+        self._writemethod(data)
+
+    def flush(self):
+        return
+
+
+if win32_and_ctypes:
+    TerminalWriter = Win32ConsoleWriter
+    import ctypes
+    from ctypes import wintypes
+
+    # ctypes access to the Windows console
+    STD_OUTPUT_HANDLE = -11
+    STD_ERROR_HANDLE  = -12
+    FOREGROUND_BLACK     = 0x0000 # black text
+    FOREGROUND_BLUE      = 0x0001 # text color contains blue.
+    FOREGROUND_GREEN     = 0x0002 # text color contains green.
+    FOREGROUND_RED       = 0x0004 # text color contains red.
+    FOREGROUND_WHITE     = 0x0007
+    FOREGROUND_INTENSITY = 0x0008 # text color is intensified.
+    BACKGROUND_BLACK     = 0x0000 # background color black
+    BACKGROUND_BLUE      = 0x0010 # background color contains blue.
+    BACKGROUND_GREEN     = 0x0020 # background color contains green.
+    BACKGROUND_RED       = 0x0040 # background color contains red.
+    BACKGROUND_WHITE     = 0x0070
+    BACKGROUND_INTENSITY = 0x0080 # background color is intensified.
+
+    SHORT = ctypes.c_short
+    class COORD(ctypes.Structure):
+        _fields_ = [('X', SHORT),
+                    ('Y', SHORT)]
+    class SMALL_RECT(ctypes.Structure):
+        _fields_ = [('Left', SHORT),
+                    ('Top', SHORT),
+                    ('Right', SHORT),
+                    ('Bottom', SHORT)]
+    class CONSOLE_SCREEN_BUFFER_INFO(ctypes.Structure):
+        _fields_ = [('dwSize', COORD),
+                    ('dwCursorPosition', COORD),
+                    ('wAttributes', wintypes.WORD),
+                    ('srWindow', SMALL_RECT),
+                    ('dwMaximumWindowSize', COORD)]
+
+    _GetStdHandle = ctypes.windll.kernel32.GetStdHandle
+    _GetStdHandle.argtypes = [wintypes.DWORD]
+    _GetStdHandle.restype = wintypes.HANDLE
+    def GetStdHandle(kind):
+        return _GetStdHandle(kind)
+
+    SetConsoleTextAttribute = ctypes.windll.kernel32.SetConsoleTextAttribute
+    SetConsoleTextAttribute.argtypes = [wintypes.HANDLE, wintypes.WORD]
+    SetConsoleTextAttribute.restype = wintypes.BOOL
+
+    _GetConsoleScreenBufferInfo = \
+        ctypes.windll.kernel32.GetConsoleScreenBufferInfo
+    _GetConsoleScreenBufferInfo.argtypes = [wintypes.HANDLE,
+                                ctypes.POINTER(CONSOLE_SCREEN_BUFFER_INFO)]
+    _GetConsoleScreenBufferInfo.restype = wintypes.BOOL
+    def GetConsoleInfo(handle):
+        info = CONSOLE_SCREEN_BUFFER_INFO()
+        _GetConsoleScreenBufferInfo(handle, ctypes.byref(info))
+        return info
+
+    def _getdimensions():
+        handle = GetStdHandle(STD_OUTPUT_HANDLE)
+        info = GetConsoleInfo(handle)
+        # Substract one from the width, otherwise the cursor wraps
+        # and the ending \n causes an empty line to display.
+        return info.dwSize.Y, info.dwSize.X - 1
+
+def write_out(fil, msg):
+    # XXX sometimes "msg" is of type bytes, sometimes text which
+    # complicates the situation.  Should we try to enforce unicode?
+    try:
+        # on py27 and above writing out to sys.stdout with an encoding
+        # should usually work for unicode messages (if the encoding is
+        # capable of it)
+        fil.write(msg)
+    except UnicodeEncodeError:
+        # on py26 it might not work because stdout expects bytes
+        if fil.encoding:
+            try:
+                fil.write(msg.encode(fil.encoding))
+            except UnicodeEncodeError:
+                # it might still fail if the encoding is not capable
+                pass
+            else:
+                fil.flush()
+                return
+        # fallback: escape all unicode characters
+        msg = msg.encode("unicode-escape").decode("ascii")
+        fil.write(msg)
+    fil.flush()
diff --git a/vllm/lib/python3.10/site-packages/py/_std.py b/vllm/lib/python3.10/site-packages/py/_std.py
new file mode 100644
index 0000000000000000000000000000000000000000..66adb7b0239e290e9a32b5d7447dbf48262b7cd8
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_std.py
@@ -0,0 +1,27 @@
+import sys
+import warnings
+
+
+class PyStdIsDeprecatedWarning(DeprecationWarning):
+    pass
+
+
+class Std(object):
+    """ makes top-level python modules available as an attribute,
+        importing them on first access.
+    """
+
+    def __init__(self):
+        self.__dict__ = sys.modules
+
+    def __getattr__(self, name):
+        warnings.warn("py.std is deprecated, please import %s directly" % name,
+                      category=PyStdIsDeprecatedWarning,
+                      stacklevel=2)
+        try:
+            m = __import__(name)
+        except ImportError:
+            raise AttributeError("py.std: could not import %s" % name)
+        return m
+
+std = Std()
diff --git a/vllm/lib/python3.10/site-packages/py/_vendored_packages/__init__.py b/vllm/lib/python3.10/site-packages/py/_vendored_packages/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/vllm/lib/python3.10/site-packages/py/_vendored_packages/iniconfig-1.1.1.dist-info/LICENSE b/vllm/lib/python3.10/site-packages/py/_vendored_packages/iniconfig-1.1.1.dist-info/LICENSE
new file mode 100644
index 0000000000000000000000000000000000000000..31ecdfb1dbc54fb41999867884dad52a07eac2ca
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_vendored_packages/iniconfig-1.1.1.dist-info/LICENSE
@@ -0,0 +1,19 @@
+
+  Permission is hereby granted, free of charge, to any person obtaining a copy
+  of this software and associated documentation files (the "Software"), to deal
+  in the Software without restriction, including without limitation the rights
+  to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+  copies of the Software, and to permit persons to whom the Software is
+  furnished to do so, subject to the following conditions:
+     
+  The above copyright notice and this permission notice shall be included in all
+  copies or substantial portions of the Software.
+ 
+  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+  IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+  FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+  AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+  LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+  OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+  SOFTWARE.
+
diff --git a/vllm/lib/python3.10/site-packages/py/_vendored_packages/iniconfig-1.1.1.dist-info/METADATA b/vllm/lib/python3.10/site-packages/py/_vendored_packages/iniconfig-1.1.1.dist-info/METADATA
new file mode 100644
index 0000000000000000000000000000000000000000..c078a7532fbbb182a4f327c8aa58de8bbd04fa03
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_vendored_packages/iniconfig-1.1.1.dist-info/METADATA
@@ -0,0 +1,78 @@
+Metadata-Version: 2.1
+Name: iniconfig
+Version: 1.1.1
+Summary: iniconfig: brain-dead simple config-ini parsing
+Home-page: http://github.com/RonnyPfannschmidt/iniconfig
+Author: Ronny Pfannschmidt, Holger Krekel
+Author-email: opensource@ronnypfannschmidt.de, holger.krekel@gmail.com
+License: MIT License
+Platform: unix
+Platform: linux
+Platform: osx
+Platform: cygwin
+Platform: win32
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: POSIX
+Classifier: Operating System :: Microsoft :: Windows
+Classifier: Operating System :: MacOS :: MacOS X
+Classifier: Topic :: Software Development :: Libraries
+Classifier: Topic :: Utilities
+Classifier: Programming Language :: Python
+Classifier: Programming Language :: Python :: 2
+Classifier: Programming Language :: Python :: 3
+
+iniconfig: brain-dead simple parsing of ini files
+=======================================================
+
+iniconfig is a small and simple INI-file parser module
+having a unique set of features:
+
+* tested against Python2.4 across to Python3.2, Jython, PyPy
+* maintains order of sections and entries
+* supports multi-line values with or without line-continuations
+* supports "#" comments everywhere
+* raises errors with proper line-numbers
+* no bells and whistles like automatic substitutions
+* iniconfig raises an Error if two sections have the same name.
+
+If you encounter issues or have feature wishes please report them to:
+
+    http://github.com/RonnyPfannschmidt/iniconfig/issues
+
+Basic Example
+===================================
+
+If you have an ini file like this::
+
+    # content of example.ini
+    [section1] # comment
+    name1=value1  # comment
+    name1b=value1,value2  # comment
+
+    [section2]
+    name2=
+        line1
+        line2
+
+then you can do::
+
+    >>> import iniconfig
+    >>> ini = iniconfig.IniConfig("example.ini")
+    >>> ini['section1']['name1'] # raises KeyError if not exists
+    'value1'
+    >>> ini.get('section1', 'name1b', [], lambda x: x.split(","))
+    ['value1', 'value2']
+    >>> ini.get('section1', 'notexist', [], lambda x: x.split(","))
+    []
+    >>> [x.name for x in list(ini)]
+    ['section1', 'section2']
+    >>> list(list(ini)[0].items())
+    [('name1', 'value1'), ('name1b', 'value1,value2')]
+    >>> 'section1' in ini
+    True
+    >>> 'inexistendsection' in ini
+    False
+
+
diff --git a/vllm/lib/python3.10/site-packages/py/_version.py b/vllm/lib/python3.10/site-packages/py/_version.py
new file mode 100644
index 0000000000000000000000000000000000000000..3d30fbec42217209b4b1ae96616a771dedae3783
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/_version.py
@@ -0,0 +1,5 @@
+# coding: utf-8
+# file generated by setuptools_scm
+# don't change, don't track in version control
+version = '1.11.0'
+version_tuple = (1, 11, 0)
diff --git a/vllm/lib/python3.10/site-packages/py/error.pyi b/vllm/lib/python3.10/site-packages/py/error.pyi
new file mode 100644
index 0000000000000000000000000000000000000000..034eba609f19b52f0cedf3d061b4b0ad96168a33
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/error.pyi
@@ -0,0 +1,129 @@
+from typing import Any, Callable, TypeVar
+
+_T = TypeVar('_T')
+
+def checked_call(func: Callable[..., _T], *args: Any, **kwargs: Any) -> _T: ...
+class Error(EnvironmentError): ...
+class EPERM(Error): ...
+class ENOENT(Error): ...
+class ESRCH(Error): ...
+class EINTR(Error): ...
+class EIO(Error): ...
+class ENXIO(Error): ...
+class E2BIG(Error): ...
+class ENOEXEC(Error): ...
+class EBADF(Error): ...
+class ECHILD(Error): ...
+class EAGAIN(Error): ...
+class ENOMEM(Error): ...
+class EACCES(Error): ...
+class EFAULT(Error): ...
+class ENOTBLK(Error): ...
+class EBUSY(Error): ...
+class EEXIST(Error): ...
+class EXDEV(Error): ...
+class ENODEV(Error): ...
+class ENOTDIR(Error): ...
+class EISDIR(Error): ...
+class EINVAL(Error): ...
+class ENFILE(Error): ...
+class EMFILE(Error): ...
+class ENOTTY(Error): ...
+class ETXTBSY(Error): ...
+class EFBIG(Error): ...
+class ENOSPC(Error): ...
+class ESPIPE(Error): ...
+class EROFS(Error): ...
+class EMLINK(Error): ...
+class EPIPE(Error): ...
+class EDOM(Error): ...
+class ERANGE(Error): ...
+class EDEADLCK(Error): ...
+class ENAMETOOLONG(Error): ...
+class ENOLCK(Error): ...
+class ENOSYS(Error): ...
+class ENOTEMPTY(Error): ...
+class ELOOP(Error): ...
+class EWOULDBLOCK(Error): ...
+class ENOMSG(Error): ...
+class EIDRM(Error): ...
+class ECHRNG(Error): ...
+class EL2NSYNC(Error): ...
+class EL3HLT(Error): ...
+class EL3RST(Error): ...
+class ELNRNG(Error): ...
+class EUNATCH(Error): ...
+class ENOCSI(Error): ...
+class EL2HLT(Error): ...
+class EBADE(Error): ...
+class EBADR(Error): ...
+class EXFULL(Error): ...
+class ENOANO(Error): ...
+class EBADRQC(Error): ...
+class EBADSLT(Error): ...
+class EDEADLOCK(Error): ...
+class EBFONT(Error): ...
+class ENOSTR(Error): ...
+class ENODATA(Error): ...
+class ETIME(Error): ...
+class ENOSR(Error): ...
+class ENONET(Error): ...
+class ENOPKG(Error): ...
+class EREMOTE(Error): ...
+class ENOLINK(Error): ...
+class EADV(Error): ...
+class ESRMNT(Error): ...
+class ECOMM(Error): ...
+class EPROTO(Error): ...
+class EMULTIHOP(Error): ...
+class EDOTDOT(Error): ...
+class EBADMSG(Error): ...
+class EOVERFLOW(Error): ...
+class ENOTUNIQ(Error): ...
+class EBADFD(Error): ...
+class EREMCHG(Error): ...
+class ELIBACC(Error): ...
+class ELIBBAD(Error): ...
+class ELIBSCN(Error): ...
+class ELIBMAX(Error): ...
+class ELIBEXEC(Error): ...
+class EILSEQ(Error): ...
+class ERESTART(Error): ...
+class ESTRPIPE(Error): ...
+class EUSERS(Error): ...
+class ENOTSOCK(Error): ...
+class EDESTADDRREQ(Error): ...
+class EMSGSIZE(Error): ...
+class EPROTOTYPE(Error): ...
+class ENOPROTOOPT(Error): ...
+class EPROTONOSUPPORT(Error): ...
+class ESOCKTNOSUPPORT(Error): ...
+class ENOTSUP(Error): ...
+class EOPNOTSUPP(Error): ...
+class EPFNOSUPPORT(Error): ...
+class EAFNOSUPPORT(Error): ...
+class EADDRINUSE(Error): ...
+class EADDRNOTAVAIL(Error): ...
+class ENETDOWN(Error): ...
+class ENETUNREACH(Error): ...
+class ENETRESET(Error): ...
+class ECONNABORTED(Error): ...
+class ECONNRESET(Error): ...
+class ENOBUFS(Error): ...
+class EISCONN(Error): ...
+class ENOTCONN(Error): ...
+class ESHUTDOWN(Error): ...
+class ETOOMANYREFS(Error): ...
+class ETIMEDOUT(Error): ...
+class ECONNREFUSED(Error): ...
+class EHOSTDOWN(Error): ...
+class EHOSTUNREACH(Error): ...
+class EALREADY(Error): ...
+class EINPROGRESS(Error): ...
+class ESTALE(Error): ...
+class EUCLEAN(Error): ...
+class ENOTNAM(Error): ...
+class ENAVAIL(Error): ...
+class EISNAM(Error): ...
+class EREMOTEIO(Error): ...
+class EDQUOT(Error): ...
diff --git a/vllm/lib/python3.10/site-packages/py/path.pyi b/vllm/lib/python3.10/site-packages/py/path.pyi
new file mode 100644
index 0000000000000000000000000000000000000000..1ddab9601ea43988162102407172f519c8cf1879
--- /dev/null
+++ b/vllm/lib/python3.10/site-packages/py/path.pyi
@@ -0,0 +1,197 @@
+from typing import Any, AnyStr, Callable, ContextManager, Generic, IO, Iterable, Iterator, List, Optional, Text, Type, Union
+from typing_extensions import Final, Literal
+import os
+import sys
+
+class _FNMatcher(Generic[AnyStr]):
+    pattern: AnyStr = ...
+    def __init__(self, pattern: AnyStr) -> None: ...
+    def __call__(self, path: local) -> bool: ...
+
+class _Stat:
+    path: Final[local] = ...
+    mode: Final[int]
+    ino: Final[int]
+    dev: Final[int]
+    nlink: Final[int]
+    uid: Final[int]
+    gid: Final[int]
+    size: Final[int]
+    atime: Final[float]
+    mtime: Final[float]
+    ctime: Final[float]
+    atime_ns: Final[int]
+    mtime_ns: Final[int]
+    ctime_ns: Final[int]
+    if sys.version_info >= (3, 8) and sys.platform == "win32":
+        reparse_tag: Final[int]
+    blocks: Final[int]
+    blksize: Final[int]
+    rdev: Final[int]
+    flags: Final[int]
+    gen: Final[int]
+    birthtime: Final[int]
+    rsize: Final[int]
+    creator: Final[int]
+    type: Final[int]
+    if sys.platform != 'win32':
+        @property
+        def owner(self) -> str: ...
+        @property
+        def group(self) -> str: ...
+    def isdir(self) -> bool: ...
+    def isfile(self) -> bool: ...
+    def islink(self) -> bool: ...
+
+
+if sys.version_info >= (3, 6):
+    _PathLike = os.PathLike
+else:
+    class _PathLike(Generic[AnyStr]):
+        def __fspath__(self) -> AnyStr: ...
+_PathType = Union[bytes, Text, _PathLike[str], _PathLike[bytes], local]
+
+class local(_PathLike[str]):
+    class ImportMismatchError(ImportError): ...
+
+    sep: Final[str]
+    strpath: Final[str]
+
+    def __init__(self, path: _PathType = ..., expanduser: bool = ...) -> None: ...
+    def __hash__(self) -> int: ...
+    def __eq__(self, other: object) -> bool: ...
+    def __ne__(self, other: object) -> bool: ...
+    def __lt__(self, other: object) -> bool: ...
+    def __gt__(self, other: object) -> bool: ...
+    def __add__(self, other: object) -> local: ...
+    def __cmp__(self, other: object) -> int: ...
+    def __div__(self, other: _PathType) -> local: ...
+    def __truediv__(self, other: _PathType) -> local: ...
+    def __fspath__(self) -> str: ...
+
+    @classmethod
+    def get_temproot(cls) -> local: ...
+    @classmethod
+    def make_numbered_dir(
+        cls,
+        prefix: str = ...,
+        rootdir: Optional[local] = ...,
+        keep: Optional[int] = ...,
+        lock_timeout: int = ...,
+    ) -> local: ...
+    @classmethod
+    def mkdtemp(cls, rootdir: Optional[local] = ...) -> local: ...
+    @classmethod
+    def sysfind(
+        cls,
+        name: _PathType,
+        checker: Optional[Callable[[local], bool]] = ...,
+        paths: Optional[Iterable[_PathType]] = ...,
+    ) -> Optional[local]: ...
+
+    @property
+    def basename(self) -> str: ...
+    @property
+    def dirname(self) -> str: ...
+    @property
+    def purebasename(self) -> str: ...
+    @property
+    def ext(self) -> str: ...
+
+    def as_cwd(self) -> ContextManager[Optional[local]]: ...
+    def atime(self) -> float: ...
+    def bestrelpath(self, dest: local) -> str: ...
+    def chdir(self) -> local: ...
+    def check(
+        self,
+        *,
+        basename: int = ..., notbasename: int = ...,
+        basestarts: int = ..., notbasestarts: int = ...,
+        dir: int = ..., notdir: int = ...,
+        dotfile: int = ..., notdotfile: int = ...,
+        endswith: int = ..., notendswith: int = ...,
+        exists: int = ..., notexists: int = ...,
+        ext: int = ..., notext: int = ...,
+        file: int = ..., notfile: int = ...,
+        fnmatch: int = ..., notfnmatch: int = ...,
+        link: int = ..., notlink: int = ...,
+        relto: int = ..., notrelto: int = ...,
+     ) -> bool: ...
+    def chmod(self, mode: int, rec: Union[int, str, Text, Callable[[local], bool]] = ...) -> None: ...
+    if sys.platform != 'win32':
+        def chown(self, user: Union[int, str], group: Union[int, str], rec: int = ...) -> None: ...
+    def common(self, other: local) -> Optional[local]: ...
+    def computehash(self, hashtype: str = ..., chunksize: int = ...) -> str: ...
+    def copy(self, target: local, mode: bool = ..., stat: bool = ...) -> None: ...
+    def dirpath(self, *args: _PathType, abs: int = ...) -> local: ...
+    def dump(self, obj: Any, bin: Optional[int] = ...) -> None: ...
+    def ensure(self, *args: _PathType, dir: int = ...) -> local: ...
+    def ensure_dir(self, *args: _PathType) -> local: ...
+    def exists(self) -> bool: ...
+    def fnmatch(self, pattern: str): _FNMatcher
+    def isdir(self) -> bool: ...
+    def isfile(self) -> bool: ...
+    def islink(self) -> bool: ...
+    def join(self, *args: _PathType, abs: int = ...) -> local: ...
+    def listdir(
+        self,
+        fil: Optional[Union[str, Text, Callable[[local], bool]]] = ...,
+        sort: Optional[bool] = ...,
+    ) -> List[local]: ...
+    def load(self) -> Any: ...
+    def lstat(self) -> _Stat: ...
+    def mkdir(self, *args: _PathType) -> local: ...
+    if sys.platform != 'win32':
+        def mklinkto(self, oldname: Union[str, local]) -> None: ...
+        def mksymlinkto(self, value: local, absolute: int = ...) -> None: ...
+    def move(self, target: local) -> None: ...
+    def mtime(self) -> float: ...
+    def new(
+        self,
+        *,
+        drive: str = ...,
+        dirname: str = ...,
+        basename: str = ...,
+        purebasename: str = ...,
+        ext: str = ...,
+    ) -> local: ...
+    def open(self, mode: str = ..., ensure: bool = ..., encoding: Optional[str] = ...) -> IO[Any]: ...
+    def parts(self, reverse: bool = ...) -> List[local]: ...
+    def pyimport(
+        self,
+        modname: Optional[str] = ...,
+        ensuresyspath: Union[bool, Literal["append", "importlib"]] = ...,
+    ) -> Any: ...
+    def pypkgpath(self) -> Optional[local]: ...
+    def read(self, mode: str = ...) -> Union[Text, bytes]: ...
+    def read_binary(self) -> bytes: ...
+    def read_text(self, encoding: str) -> Text: ...
+    def readlines(self, cr: int = ...) -> List[str]: ...
+    if sys.platform != 'win32':
+        def readlink(self) -> str: ...
+    def realpath(self) -> local: ...
+    def relto(self, relpath: Union[str, local]) -> str: ...
+    def remove(self, rec: int = ..., ignore_errors: bool = ...) -> None: ...
+    def rename(self, target: _PathType) -> None: ...
+    def samefile(self, other: _PathType) -> bool: ...
+    def setmtime(self, mtime: Optional[float] = ...) -> None: ...
+    def size(self) -> int: ...
+    def stat(self, raising: bool = ...) -> _Stat: ...
+    def sysexec(self, *argv: Any, **popen_opts: Any) -> Text: ...
+    def visit(
+        self,
+        fil: Optional[Union[str, Text, Callable[[local], bool]]] = ...,
+        rec: Optional[Union[Literal[1, True], str, Text, Callable[[local], bool]]] = ...,
+        ignore: Type[Exception] = ...,
+        bf: bool = ...,
+        sort: bool = ...,
+    ) -> Iterator[local]: ...
+    def write(self, data: Any, mode: str = ..., ensure: bool = ...) -> None: ...
+    def write_binary(self, data: bytes, ensure: bool = ...) -> None: ...
+    def write_text(self, data: Union[str, Text], encoding: str, ensure: bool = ...) -> None: ...
+
+
+# Untyped types below here.
+svnwc: Any
+svnurl: Any
+SvnAuth: Any
diff --git a/vllm/lib/python3.10/site-packages/py/py.typed b/vllm/lib/python3.10/site-packages/py/py.typed
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/__pycache__/debug.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/__pycache__/debug.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..24e8912b4ce326f6e1e836858c4aac84f0834d13
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/__pycache__/debug.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/__init__.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..06d81adb33cdb88f9938727ca3920d2fdde71e50
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/__init__.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/_virtualenv.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/_virtualenv.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9dc861af48ff78613c3fa41f65c9b69c210cd068
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/_virtualenv.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/api.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/api.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ce7cd31292db4be9d7812ab9f1a49d5645158a49
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/api.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/store.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/store.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6cf9ba52de3edc35e5078cbde60c4d586f7c1f98
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/store.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/venv.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/venv.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..cb5eb181a60971f0e54f24fed2abf7b2159c897b
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/__pycache__/venv.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/builtin/__pycache__/ref.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/builtin/__pycache__/ref.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8febb7cd3bf9149a854e350293cb52b386db7905
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/builtin/__pycache__/ref.cpython-310.pyc differ
diff --git a/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/builtin/__pycache__/via_global_self_do.cpython-310.pyc b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/builtin/__pycache__/via_global_self_do.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..cb9fd1df811645f86b93d97e4ec3f581296b39f1
Binary files /dev/null and b/vllm/lib/python3.10/site-packages/virtualenv/create/via_global_ref/builtin/__pycache__/via_global_self_do.cpython-310.pyc differ